Examenos

49.1% · raw 49.1 %

Headline · xhigh effort · Own vendor

Source/record date: 2026-11-09

Anthropic Claude Fable 5.2 launch table (9 November 2026); safeguards on, highest available effort unless noted. Best effort; Cognition ran the evaluation (Claude Code harness), mean@5 overall score.

34.2% · raw 34.2 %

Alternative · low effort · Own vendor

Source/record date: 2026-11-09

Anthropic Claude Fable 5.2 and Mythos 5.2 system card, p. 209. Cognition run, Claude Code harness; effort sweep stated in text.

37.6% · raw 37.6 %

Alternative · medium effort · Own vendor

Source/record date: 2026-11-09

Anthropic Claude Fable 5.2 and Mythos 5.2 system card, p. 209. Cognition run, Claude Code harness; effort sweep stated in text.

45.3% · raw 45.3 %

Alternative · high effort · Own vendor

Source/record date: 2026-11-09

Anthropic Claude Fable 5.2 and Mythos 5.2 system card, p. 209. Cognition run, Claude Code harness; effort sweep stated in text.

45.7% · raw 45.7 %

Alternative · max effort · Own vendor

Source/record date: 2026-11-09

Anthropic Claude Fable 5.2 and Mythos 5.2 system card, p. 209. Cognition run; includes graded scores of 22 of 660 timed-out attempts (44.4% if counted as zero).

34.2% · raw 34.2 %

Alternative · low effort · Own vendor

Source/record date: 2026-11-09

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.3742, "tokens": 23907, "cost": 2.623, "flagged_rate": 0.01515}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 2026-11-10; evaluation date unpublished. Benchmark-owner board row; the headline is Anthropic's launch value (49.1%, xhigh), which matches this board within rounding.

37.6% · raw 37.59 %

Alternative · medium effort · Own vendor

Source/record date: 2026-11-09

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.4106, "tokens": 28384, "cost": 3.103, "flagged_rate": 0.01667}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 2026-11-10; evaluation date unpublished. Benchmark-owner board row; the headline is Anthropic's launch value (49.1%, xhigh), which matches this board within rounding.

45.3% · raw 45.29 %

Alternative · high effort · Own vendor

Source/record date: 2026-11-09

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.4939, "tokens": 75978, "cost": 8.402, "flagged_rate": 0.0197}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 2026-11-10; evaluation date unpublished. Benchmark-owner board row; the headline is Anthropic's launch value (49.1%, xhigh), which matches this board within rounding.

49.1% · raw 49.11 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-11-09

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.5348, "tokens": 156009, "cost": 17.734, "flagged_rate": 0.01515}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 2026-11-10; evaluation date unpublished. Benchmark-owner board row; the headline is Anthropic's launch value (49.1%, xhigh), which matches this board within rounding.

45.7% · raw 45.65 %

Alternative · max effort · Own vendor

Source/record date: 2026-11-09

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.5, "tokens": 417808, "cost": 55.868, "flagged_rate": 0.03939}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 2026-11-10; evaluation date unpublished. Benchmark-owner board row; the headline is Anthropic's launch value (49.1%, xhigh), which matches this board within rounding.

14.7% · raw 14.72 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.1636, "tokens": 14230, "cost": 0.746, "flagged_rate": 0.0121}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

30.1% · raw 30.1 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.3303, "tokens": 30068, "cost": 1.699, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

34.5% · raw 34.53 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.3788, "tokens": 41233, "cost": 2.423, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

43.4% · raw 43.38 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.4727, "tokens": 88053, "cost": 5.184, "flagged_rate": 0.0212}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

45.5% · raw 45.45 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.497, "tokens": 201919, "cost": 12.821, "flagged_rate": 0.0288}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

31.1% · raw 31.11 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.3424, "tokens": 8987, "cost": 2.437, "flagged_rate": 0.0076}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

36.8% · raw 36.75 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.4015, "tokens": 14065, "cost": 3.493, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

38.3% · raw 38.31 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.4136, "tokens": 18433, "cost": 4.277, "flagged_rate": 0.0242}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

41.4% · raw 41.41 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.4485, "tokens": 20869, "cost": 4.539, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

42.8% · raw 42.77 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.4621, "tokens": 36354, "cost": 6.676, "flagged_rate": 0.0227}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

9.4% · raw 9.41 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.1062, "tokens": 11265, "cost": 0.369, "flagged_rate": 0.0106}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

14.2% · raw 14.17 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.1576, "tokens": 14449, "cost": 0.482, "flagged_rate": 0.0136}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

21.9% · raw 21.9 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.2428, "tokens": 27839, "cost": 1.066, "flagged_rate": 0.0137}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

34.8% · raw 34.75 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.3803, "tokens": 80433, "cost": 3.315, "flagged_rate": 0.0227}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

41% · raw 40.96 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.45, "tokens": 514722, "cost": 19.941, "flagged_rate": 0.0379}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

31.9% · raw 31.88 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-28

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.3535, "tokens": 57868, "cost": 4.548, "flagged_rate": 0.023}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 1 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board changelog dates this addition 28 October.

40% · raw 40 %

Headline · medium effort · Own vendor

Source/record date: 2026-10-28

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.4432, "tokens": 109126, "cost": 4.044, "flagged_rate": 0.0105}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 1 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board changelog dates this addition 28 October.

37.5% · raw 37.48 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-28

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.4167, "tokens": 121442, "cost": 4.434, "flagged_rate": 0.0183}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 1 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board changelog dates this addition 28 October.

41.1% · raw 41.1 %

Alternative · medium effort · Peer vendor

Source/record date: 2026-11-09

As reported by Anthropic in the Claude Fable 5.2 launch table (9 November 2026). Medium effort as labelled in the table.

25.6% · raw 25.6 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.2833, "tokens": 10042, "cost": 0.341, "flagged_rate": 0.0091}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

29.6% · raw 29.58 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.3227, "tokens": 15686, "cost": 0.492, "flagged_rate": 0.0136}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

34.8% · raw 34.82 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.3803, "tokens": 22419, "cost": 0.647, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

32.3% · raw 32.33 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.353, "tokens": 24789, "cost": 0.695, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

39% · raw 38.95 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.4227, "tokens": 41479, "cost": 1.029, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

20.6% · raw 20.59 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.2288, "tokens": 28797, "cost": 3.007, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

21.5% · raw 21.47 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.2379, "tokens": 37911, "cost": 3.891, "flagged_rate": 0.0182}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

25.7% · raw 25.67 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.2848, "tokens": 54242, "cost": 5.386, "flagged_rate": 0.0182}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

32.4% · raw 32.44 %

Headline · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.3576, "tokens": 92813, "cost": 8.68, "flagged_rate": 0.0212}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

31.7% · raw 31.67 %

Alternative · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.3485, "tokens": 119089, "cost": 11.457, "flagged_rate": 0.0197}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

20.8% · raw 20.8 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: grok-build. {"correct": 0.2337, "tokens": 38161, "cost": 3.873, "flagged_rate": 0.0106}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

25.9% · raw 25.94 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: grok-build. {"correct": 0.2924, "tokens": 72180, "cost": 8.028, "flagged_rate": 0.0091}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

26.6% · raw 26.55 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: grok-build. {"correct": 0.297, "tokens": 94341, "cost": 11.094, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

28.6% · raw 28.62 %

Headline · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: grok-build. {"correct": 0.3197, "tokens": 110790, "cost": 14.36, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

10.5% · raw 10.46 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.1168, "tokens": 39366, "cost": 0.065, "flagged_rate": 0.0091}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

13.4% · raw 13.42 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.15, "tokens": 57582, "cost": 0.097, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

18.6% · raw 18.64 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.2076, "tokens": 80993, "cost": 0.139, "flagged_rate": 0.0197}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

22.3% · raw 22.28 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.2477, "tokens": 127735, "cost": 0.226, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

27.7% · raw 27.72 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: claude-code. {"correct": 0.305, "tokens": 205112, "cost": 0.366, "flagged_rate": 0.0258}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

34.9% · raw 34.9 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “Claude Haiku 5.5”, value “34.9%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

21.5% · raw 21.47 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.2379, "tokens": 63291, "cost": 3.977, "flagged_rate": 0.0136}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

23.5% · raw 23.45 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.2606, "tokens": 123879, "cost": 5.688, "flagged_rate": 0.0091}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

26.3% · raw 26.3 %

Headline · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.2909, "tokens": 148569, "cost": 6.514, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

25.4% · raw 25.4 %

Headline · xhigh effort · Own vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “Inkling 2 Small”, value “25.4%”. Thinking Machines states it uses Artificial Analysis’s reported score for this benchmark, for its own and other models; not an Artificial Analysis record captured by Examenos. Artificial Analysis’s run of Cognition’s FrontierCode 2.0; Cognition’s own board has no Inkling 2 Small entry at the check. Thinking Machines runs its evaluations at effort 0.99, which the post labels “xhigh”, temperature 1.0 and a 1M-token maximum trajectory (256K for Inkling-Small); stored as xhigh because the vendor names it so. Execution date unpublished; the table calls itself an evaluation snapshot.

21.2% · raw 21.21 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: muse. {"correct": 0.2413, "tokens": 20301, "cost": 1.843, "flagged_rate": 0.0152}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: Meta's model docs name the Model API variant "Muse Wonder (muse-wonder-1.0)"; Muse Wonder Full has no API ID and is a separate variant, so this is Muse Wonder, not Wonder Full.

23.1% · raw 23.14 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: muse. {"correct": 0.2606, "tokens": 26714, "cost": 2.534, "flagged_rate": 0.0182}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: Meta's model docs name the Model API variant "Muse Wonder (muse-wonder-1.0)"; Muse Wonder Full has no API ID and is a separate variant, so this is Muse Wonder, not Wonder Full.

24.3% · raw 24.28 %

Headline · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: muse. {"correct": 0.2742, "tokens": 32832, "cost": 3.143, "flagged_rate": 0.0167}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: Meta's model docs name the Model API variant "Muse Wonder (muse-wonder-1.0)"; Muse Wonder Full has no API ID and is a separate variant, so this is Muse Wonder, not Wonder Full.

24.1% · raw 24.11 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: muse. {"correct": 0.2747, "tokens": 38672, "cost": 3.793, "flagged_rate": 0.0197}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: Meta's model docs name the Model API variant "Muse Wonder (muse-wonder-1.0)"; Muse Wonder Full has no API ID and is a separate variant, so this is Muse Wonder, not Wonder Full.

24% · raw 24 %

Headline · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “MiMo V2.6 Pro”, value “24.0%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

23.2% · raw 23.17 %

Alternative · low effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin. {"cost": 0.203}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Cognition’s first 5 November copy of the file also listed a high-effort run (18.77, cost 0.2185); the current file has withdrawn it, so it is not stored.

24% · raw 24 %

Headline · max effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin. {"cost": 0.2024}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Cognition’s first 5 November copy of the file also listed a high-effort run (18.77, cost 0.2185); the current file has withdrawn it, so it is not stored.

15.6% · raw 15.56 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin. {"correct": 0.176, "tokens": 28705, "cost": 0.442, "flagged_rate": 0.0197}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

23.1% · raw 23.06 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin. {"correct": 0.2595, "tokens": 68546, "cost": 0.958, "flagged_rate": 0.0288}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

23.8% · raw 23.78 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin. {"correct": 0.2667, "tokens": 97584, "cost": 1.261, "flagged_rate": 0.0242}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

4.1% · raw 4.05 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.0455, "tokens": 6519, "cost": 0.023, "flagged_rate": 0.003}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

16% · raw 15.97 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.1773, "tokens": 31360, "cost": 0.092, "flagged_rate": 0.0015}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

16.5% · raw 16.5 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.1848, "tokens": 42170, "cost": 0.115, "flagged_rate": 0.003}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

18.4% · raw 18.44 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.2045, "tokens": 50565, "cost": 0.132, "flagged_rate": 0.0045}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

21.9% · raw 21.87 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: codex. {"correct": 0.2424, "tokens": 79659, "cost": 0.174, "flagged_rate": 0.0121}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

25.2% · raw 25.2 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “GPT-6 Luna”, value “25.2%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

21.6% · raw 21.6 %

Headline · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “MiMo V2.6 Flash”, value “21.6%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

20.8% · raw 20.8 %

Headline · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “MiniMax-M3.1-Flash”, value “20.8%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

19.7% · raw 19.7 %

Alternative · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights. Cognition FrontierCode 2.0, as charted by Mistral. Cognition’s board had no Mistral Large 4 row at the 29 October check. Demoted 2026-11-01: duplicate of headline from Cognition board; the board row now exists at 19.65 (high, Mistral Vibe harness).

9.5% · raw 9.5 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: vibe. {"correct": 0.1061, "tokens": 28384, "cost": 0.669, "flagged_rate": 0.0}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 1 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board label Mistral Large 4; Mistral charts about 19.7 for the released weights, which matches the high run. The board changelog lists this replacement under 27 October, while the row first appeared in the data after the 29 October check.

8.8% · raw 8.78 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: vibe. {"correct": 0.0985, "tokens": 51238, "cost": 1.031, "flagged_rate": 0.0}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 1 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board label Mistral Large 4; Mistral charts about 19.7 for the released weights, which matches the high run. The board changelog lists this replacement under 27 October, while the row first appeared in the data after the 29 October check.

19.7% · raw 19.65 %

Headline · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: vibe. {"correct": 0.2197, "tokens": 88069, "cost": 1.549, "flagged_rate": 0.0}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 1 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board label Mistral Large 4; Mistral charts about 19.7 for the released weights, which matches the high run. The board changelog lists this replacement under 27 October, while the row first appeared in the data after the 29 October check.

15% · raw 15 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “Mistral Large 4”, value “15.0%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. The table lists 49B active / 1050B total for this column; Mistral publishes 52B active / 1.05T total. Peer effort and run dates are not published, so effort is unknown.

12.9% · raw 12.86 %

Alternative · low effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1414, "tokens": 20367, "cost": 1.877, "flagged_rate": 0.0}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Board changelog: “Nov 5, 2026 — Added Kimi K3.1 Pro.” Data checked 7 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

17.3% · raw 17.29 %

Alternative · high effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1919, "tokens": 27628, "cost": 2.358, "flagged_rate": 0.0076}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Board changelog: “Nov 5, 2026 — Added Kimi K3.1 Pro.” Data checked 7 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

19.6% · raw 19.64 %

Headline · max effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.2172, "tokens": 37242, "cost": 3.033, "flagged_rate": 0.0051}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Board changelog: “Nov 5, 2026 — Added Kimi K3.1 Pro.” Data checked 7 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

16% · raw 15.96 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.182, "tokens": 40659, "cost": 0.206, "flagged_rate": 0.0265}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: the label equals DeepSeek's September checkpoint name; Cognition added it on 13 October, before the 1017 refresh, and labels DeepSeek checkpoints with date suffixes (V4 Pro 0813, V4 Flash 0731). The 2.0 run date is unpublished; if Cognition re-ran it through the moving deepseek-flash alias after 17 October, it would be 1017 instead. Identity corrected 4 November 2026: Cognition’s data.json now labels these runs “DeepSeek V4.1 Flash 1017” (same values as the earlier “DeepSeek V4.1 Flash” label), so they belong to the 1017 checkpoint.

18.2% · raw 18.21 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.2062, "tokens": 57169, "cost": 0.274, "flagged_rate": 0.0286}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: the label equals DeepSeek's September checkpoint name; Cognition added it on 13 October, before the 1017 refresh, and labels DeepSeek checkpoints with date suffixes (V4 Pro 0813, V4 Flash 0731). The 2.0 run date is unpublished; if Cognition re-ran it through the moving deepseek-flash alias after 17 October, it would be 1017 instead. Identity corrected 4 November 2026: Cognition’s data.json now labels these runs “DeepSeek V4.1 Flash 1017” (same values as the earlier “DeepSeek V4.1 Flash” label), so they belong to the 1017 checkpoint.

19.3% · raw 19.3 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.2177, "tokens": 82220, "cost": 0.384, "flagged_rate": 0.0228}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Identity: the label equals DeepSeek's September checkpoint name; Cognition added it on 13 October, before the 1017 refresh, and labels DeepSeek checkpoints with date suffixes (V4 Pro 0813, V4 Flash 0731). The 2.0 run date is unpublished; if Cognition re-ran it through the moving deepseek-flash alias after 17 October, it would be 1017 instead. Identity corrected 4 November 2026: Cognition’s data.json now labels these runs “DeepSeek V4.1 Flash 1017” (same values as the earlier “DeepSeek V4.1 Flash” label), so they belong to the 1017 checkpoint.

28.5% · raw 28.5 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “DeepSeek-V4.1-Flash-1017”, value “28.5%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

18.7% · raw 18.7 %

Headline · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “Gemini 3.8 Flash”, value “18.7%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

14.4% · raw 14.37 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1616, "tokens": 47199, "cost": 3.189, "flagged_rate": 0.0177}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

16.8% · raw 16.84 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1894, "tokens": 73247, "cost": 5.181, "flagged_rate": 0.0227}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

17.8% · raw 17.81 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.202, "tokens": 100706, "cost": 7.516, "flagged_rate": 0.0429}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

21.6% · raw 21.6 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “GLM 5.3”, value “21.6%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

11.1% · raw 11.1 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1237, "tokens": 28740, "cost": 2.977, "flagged_rate": 0.0177}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

17.7% · raw 17.74 %

Headline · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1975, "tokens": 49854, "cost": 4.475, "flagged_rate": 0.0203}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

17.3% · raw 17.31 %

Alternative · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1944, "tokens": 64271, "cost": 5.456, "flagged_rate": 0.0202}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

17.3% · raw 17.3 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “Kimi K3”, value “17.3%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

7.8% · raw 7.81 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0888, "tokens": 35321, "cost": 0.43, "flagged_rate": 0.0254}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

12.4% · raw 12.41 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1414, "tokens": 56914, "cost": 0.584, "flagged_rate": 0.0253}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

14.2% · raw 14.18 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.1587, "tokens": 90044, "cost": 0.792, "flagged_rate": 0.0234}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

14.2% · raw 14.2 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “GLM-5.3-Flash”, value “14.2%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

0.4% · raw 0.42 %

Alternative · low effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"cost": 0.593}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

1.6% · raw 1.63 %

Alternative · medium effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"cost": 1.865}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

4% · raw 3.95 %

Alternative · high effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"cost": 6.78}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

6.2% · raw 6.24 %

Headline · xhigh effort · Own vendor

Source/record date: 2026-11-04

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"cost": 9.157}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

7.6% · raw 7.6 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-11-05

Thinking Machines’ Introducing Inkling 2 Small (post and model card, 5 November 2026), benchmark table, row “FrontierCode v2.0”, column “Muse Spark Open”, value “7.6%”. As reported by Thinking Machines; not a result from the model’s developer. Thinking Machines attributes this row to Artificial Analysis. Peer effort and run dates are not published, so effort is unknown.

1.3% · raw 1.27 %

Alternative · high effort · Own vendor

Source/record date: 2026-07-30

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0154, "tokens": 13658, "cost": 1.366, "flagged_rate": 0.0256}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board effort labels for Inkling-Small map Thinking Machines’ 0–1 setting to Cognition’s names; Cognition does not publish the numeric setting.

3.2% · raw 3.21 %

Headline · max effort · Own vendor

Source/record date: 2026-07-30

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0379, "tokens": 39077, "cost": 3.684, "flagged_rate": 0.0606}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 5 November; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it. Board effort labels for Inkling-Small map Thinking Machines’ 0–1 setting to Cognition’s names; Cognition does not publish the numeric setting.

0% · raw 0 %

Alternative · low effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0, "tokens": 3490, "cost": 0.142, "flagged_rate": 0.0051}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

0.8% · raw 0.83 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0101, "tokens": 7243, "cost": 0.315, "flagged_rate": 0.0127}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

1.7% · raw 1.74 %

Alternative · high effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0202, "tokens": 14068, "cost": 0.612, "flagged_rate": 0.0227}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

2.4% · raw 2.43 %

Alternative · xhigh effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0278, "tokens": 23156, "cost": 1.126, "flagged_rate": 0.0379}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.

2.9% · raw 2.9 %

Headline · max effort · Own vendor

Source/record date: 2026-10-27

Cognition primary public data.json (https://cognition.com/data/frontier-code-2.0/data.json), FrontierCode 2.0 main, 132 tasks; weighted new_score x100, not unweighted correct. Harness: devin-cli. {"correct": 0.0329, "tokens": 27971, "cost": 1.299, "flagged_rate": 0.0278}. Cost is source-observed mean USD per rollout at list API pricing, not an API rate. Data checked 28 October; evaluation date unpublished. Headline is the best-scoring published effort, as Cognition displays it.