Examenos

19.7% · raw 19.7 %

Headline · 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.

41.4% · raw 41.4 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights. Chart title “Artificial Analysis Coding Agent Index - Terminal-Bench 4.0”; the post’s text rounds to 41%.

50.9% · raw 50.9 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights. Artificial Analysis Cyber Index as charted by Mistral; the post’s text rounds to 51%.

96.6% · raw 96.6 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights. Average refusal on the cyber-related prompts of JailbreakBench, StrongREJECT and AgentHarm.

73.3% · raw 73.3 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights.

84.5% · raw 84.5 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights.

97.1% · raw 97.1 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s 29 October chart for the released weights.

48% · raw 48 %

Headline · high effort · Own vendor

Source/record date: 2026-10-29

Mistral’s chart of Artificial Analysis Coding Agent Index scores, “Mistral Large 4 - High”. The model card also cites 48%.

35% · raw 35 %

Alternative · medium effort · Own vendor

Source/record date: 2026-10-29

Mistral’s chart of Artificial Analysis Coding Agent Index scores, “Mistral Large 4 - Medium”.

26 · raw 26 index

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Artificial Analysis Intelligence Index v5 as cited on Mistral’s model card.

64% · raw 64 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s Surge Code chart: share of prompts where graders preferred Mistral Large 4 over GLM-5.3.

64% · raw 64 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s Surge Code chart: share of prompts where graders preferred Mistral Large 4 over Kimi K3.

56% · raw 56 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s Surge Code chart: share of prompts where graders preferred Mistral Large 4 over DeepSeek V4.1 Flash.

22% · raw 22 %

Headline · unknown effort · Own vendor

Source/record date: 2026-10-29

Mistral’s Surge Code chart: share of prompts where graders preferred Mistral Large 4 over Claude Opus 5.5.

48 · raw 48 index

Headline · high effort · Independent evaluator

Source/record date: 2026-10-29

Harness: Mistral Vibe. Model-plus-agent measurement, not a bare-model score. AA labels this Mistral Large 4, not Preview, and Mistral's 29 October post gives 48% to the released weights. Moved 2026-11-01 from mistral-large-4. Manual observation of Artificial Analysis's public Coding Agents page, "Coding Agent Index v2.0" chart, 29 October 2026 (all 30 entries selected; rendered bar label paired with the x-axis entry label; research-raw/aa-v5-2026-10-29/cap_coding_agent_v2.txt). Configuration "Mistral Vibe - Mistral Large 4 (High)". AA's displayed rounding; no API, download, hidden precision or inferred run date.

35 · raw 35 index

Alternative · medium effort · Independent evaluator

Source/record date: 2026-10-29

Harness: Mistral Vibe. Model-plus-agent measurement, not a bare-model score. AA labels this Mistral Large 4, not Preview, and Mistral's 29 October post gives 48% to the released weights. Moved 2026-11-01 from mistral-large-4. Manual observation of Artificial Analysis's public Coding Agents page, "Coding Agent Index v2.0" chart, 29 October 2026 (all 30 entries selected; rendered bar label paired with the x-axis entry label; research-raw/aa-v5-2026-10-29/cap_coding_agent_v2.txt). Configuration "Mistral Vibe - Mistral Large 4 (Medium)". AA's displayed rounding; no API, download, hidden precision or inferred run date.

56.420% · raw 56.42 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Vals Index, 56.42%, published uncertainty ±1.26 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

56.180% · raw 56.18 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Vals Finance Index, 56.18%, published uncertainty ±1.40 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

67.680% · raw 67.68 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Vals Legal Index, 67.68%, published uncertainty ±1.70 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

68.670% · raw 68.67 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label EMB v1.1, 68.67%, published uncertainty ±2.43 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

61.720% · raw 61.72 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Finance Agent v2.1, 61.72%, published uncertainty ±2.05 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

35.580% · raw 35.58 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Legal Research Bench, 35.58%, published uncertainty ±3.33 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

27.330% · raw 27.33 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Tax Agent Bench, 27.33%, published uncertainty ±2.95 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

81.560% · raw 81.56 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Vibe Code Bench v1.1, 81.56%, published uncertainty ±3.29 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished. Harness: OpenHands.

15.830% · raw 15.83 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label HLAB-Vals-1.1, 15.83%, published uncertainty ±3.13 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished.

26.260% · raw 26.26 %

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Vals exact model page https://www.vals.ai/models/mistralai_mistral-large-4-oss; source label Terminal-Bench 4.0, 26.26%, published uncertainty ±5.13 percentage points. Vals calls this model Mistral Large 4 OSS, released 29 October 2026. Page defaults: effort high, temperature 1, top_p 0.95, max-output harness ceiling 256,000, 1M context; individual suites can override defaults. Model page defaults are not inferred vendor product limits. Run execution date unpublished. Harness: Mini-SWE-agent.

$3.717/task · raw 3.717 USD/task

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

Publisher-displayed Vals Index mean cost per test; named Vals aggregate/harness only, not token pricing, subscription cost or provider entitlement. Suite run execution dates unpublished.

26 · raw 26 index

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

AA lists Mistral Large 4, Mistral Large 4 Preview and Mistral Large 4 (Medium) separately; this is “Mistral Large 4 (High)”. Mistral’s model card also cites 26. Manual observation of Artificial Analysis’s public model page, 1 November 2026. AA’s displayed rounding; no API, download, hidden precision or inferred run date.

$2.06/task · raw 2.06 USD/task

Headline · high effort · Independent evaluator

Source/record date: 2026-11-01

AA usage-priced cost per task at its list pricing; not an API tariff or invoice. Manual observation of Artificial Analysis’s public model page, 1 November 2026. AA’s displayed rounding; no API, download, hidden precision or inferred run date.