Examenos

Hy4 preview Benchmarks, Specifications & Availability

Explore Hy4 preview from Tencent: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.

Compare Hy4 preview with other models →Explore data coverage

Published specifications

Provider
Tencent
Access
Open weights
License
Apache 2.0
Context window
1M
Total parameters
770B
Active parameters
49B
Released
2026-08-28
Modalities
text
Family
Hunyuan Hy4

Model card · Announcement · Website · OpenRouter

Model notes

Open-weight MoE; vendor charts emphasize coding/office/science agents. Open weights verified on Hugging Face https://huggingface.co/tencent/Hy4-preview (2026-09-25).

Pricing · OpenRouter

Dated cached OpenRouter rates in USD per 1M tokens. Open the dashboard for live enhancements. Per-metric endpoint minima can refer to different providers; they are not a guaranteed combined rate from one endpoint.

Recorded pricing tiers
TierInput / 1MOutput / 1MCached input / 1MCache write / 1MDate & source
Default[object Object][object Object][object Object]—2026-10-03 · OpenRouter source
Recorded pricing notes

min-healthy endpoint minima 2026-10-01; time-window overrides apply on a winning provider; min-healthy endpoint minima 2026-10-01; min-healthy endpoint minima 2026-10-02; min-healthy endpoint minima 2026-10-03

Official / vendor benchmarks

Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.

Official / vendor headline scores; expand evidence for every record
Benchmark / evaluatorHeadline scoreEvidence
Agents' Last Exam22.8%unknown effort
All 2 recorded results & sources

22.8% · raw 22.8 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

ALE-CLI; vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks

22.8% · raw 22.8 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
APEX-Agents37.1%unknown effort
All 1 recorded result & sources

37.1% · raw 37.1 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

APEX-Agents pass@1

https://hy4.site/benchmarks/hy4-benchmarks
AutomationBench v1.0.632.1%unknown effort
All 1 recorded result & sources

32.1% · raw 32.1 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
BioMysteryBench71.3%unknown effort
All 1 recorded result & sources

71.3% · raw 71.3 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

BioMysteryBench

https://hy4.site/benchmarks/hy4-benchmarks
CyberGym78.4%unknown effort
All 1 recorded result & sources

78.4% · raw 78.4 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
DeepSWE v1.164.3%unknown effort
All 2 recorded results & sources

64.3% · raw 64.3 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription (DeepSWE)

https://hy4.site/benchmarks/hy4-benchmarks

64.3% · raw 64.3 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
GDPval-AA v21678unknown effort
All 1 recorded result & sources

1678 · raw 1678 Elo

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Elo; vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
GPQA Diamond92.3%unknown effort
All 1 recorded result & sources

92.3% · raw 92.3 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
HLE (text, no tools)43.4%unknown effort
All 1 recorded result & sources

43.4% · raw 43.4 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

HLE text no tools

https://hy4.site/benchmarks/hy4-benchmarks
HorizonMath (pass@4)8.8%unknown effort
All 1 recorded result & sources

8.8% · raw 8.8 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

HorizonMath pass@4

https://hy4.site/benchmarks/hy4-benchmarks
JobBench61.7%unknown effort
All 1 recorded result & sources

61.7% · raw 61.7 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
NL2Repo-Bench58.9%unknown effort
All 2 recorded results & sources

58.9% · raw 58.9 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks

58.9% · raw 58.9 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
OneMillionBench (w/ tools)65.4%unknown effort
All 1 recorded result & sources

65.4% · raw 65.4 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

OneMillionBench w/ tools

https://hy4.site/benchmarks/hy4-benchmarks
PostTrainBench35.6%unknown effort
All 2 recorded results & sources

35.6% · raw 35.6 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

PostTrainBench

https://hy4.site/benchmarks/hy4-benchmarks

35.6% · raw 35.6 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
ProgramBench17.5%unknown effort
All 2 recorded results & sources

17.5% · raw 17.5 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

ProgramBench; hy4.site transcription of Tencent vendor charts

https://hy4.site/benchmarks/hy4-benchmarks

17.5% · raw 17.5 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
SkillsBench v1.162.9%unknown effort
All 1 recorded result & sources

62.9% · raw 62.9 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

SkillsBench v1.1

https://hy4.site/benchmarks/hy4-benchmarks
SWE Atlas Refactoring53.3%unknown effort
All 1 recorded result & sources

53.3% · raw 53.3 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

SWE Atlas Refactoring

https://hy4.site/benchmarks/hy4-benchmarks
SWE-bench Multilingual82.9%unknown effort
All 1 recorded result & sources

82.9% · raw 82.9 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
SWE-bench Pro65.7%unknown effort
All 2 recorded results & sources

65.7% · raw 65.7 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks

65.7% · raw 65.7 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
Terminal-Bench 2.185.4%unknown effort
All 2 recorded results & sources

85.4% · raw 85.4 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Transcribed from Tencent vendor launch charts; official page is chart-heavy

https://hy4.site/benchmarks/hy4-benchmarks

85.4% · raw 85.4 %

Alternative · unknown effort · Peer vendor

Source/record date: 2026-09-27

As reported by NaiveAI.

https://naive.ai/en/research/
Toolathlon-Verified74.1%unknown effort
All 1 recorded result & sources

74.1% · raw 74.1 %

Headline · unknown effort · Own vendor

Source/record date: 2026-08-28

Vendor chart transcription

https://hy4.site/benchmarks/hy4-benchmarks
SWE-Atlas QnA64%unknown effort
All 1 recorded result & sources

64% · raw 64 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; SWE Atlas Codebase Q&A

https://huggingface.co/tencent/Hy4-preview
SWE Atlas Test Writing57.8%unknown effort
All 1 recorded result & sources

57.8% · raw 57.8 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; SWE Atlas Test Writing

https://huggingface.co/tencent/Hy4-preview
SWE-Marathon31.9%unknown effort
All 1 recorded result & sources

31.9% · raw 31.9 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; SWE-Marathon

https://huggingface.co/tencent/Hy4-preview
Harbor-Index39.6%unknown effort
All 1 recorded result & sources

39.6% · raw 39.6 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; Harbor-Index

https://huggingface.co/tencent/Hy4-preview
WideSearch83.9%unknown effort
All 1 recorded result & sources

83.9% · raw 83.9 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; WideSearch

https://huggingface.co/tencent/Hy4-preview
DRACO77.2%unknown effort
All 1 recorded result & sources

77.2% · raw 77.2 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; DRACO

https://huggingface.co/tencent/Hy4-preview
OfficeQA Pro66.2%unknown effort
All 1 recorded result & sources

66.2% · raw 66.2 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; OfficeQA Pro

https://huggingface.co/tencent/Hy4-preview
MCP Atlas83.7%unknown effort
All 1 recorded result & sources

83.7% · raw 83.7 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; MCP-Atlas (public)

https://huggingface.co/tencent/Hy4-preview
WorkspaceBench60.2%unknown effort
All 1 recorded result & sources

60.2% · raw 60.2 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; WorkspaceBench

https://huggingface.co/tencent/Hy4-preview
BankerToolBench78.6%unknown effort
All 1 recorded result & sources

78.6% · raw 78.6 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; BankerToolBench

https://huggingface.co/tencent/Hy4-preview
Humanity's Last Exam (w/ tools)55.4%unknown effort
All 1 recorded result & sources

55.4% · raw 55.4 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; HLE with tools, text-only

https://huggingface.co/tencent/Hy4-preview
CritPt16.9%unknown effort
All 1 recorded result & sources

16.9% · raw 16.9 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; CritPt (official)

https://huggingface.co/tencent/Hy4-preview
SUPERChem66.4%unknown effort
All 1 recorded result & sources

66.4% · raw 66.4 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; SUPERChem

https://huggingface.co/tencent/Hy4-preview
ArXivMath66.6%unknown effort
All 1 recorded result & sources

66.6% · raw 66.6 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; ArXivMath

https://huggingface.co/tencent/Hy4-preview
MathArena Apex74.2%unknown effort
All 1 recorded result & sources

74.2% · raw 74.2 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; MathArena Apex 2025

https://huggingface.co/tencent/Hy4-preview
BrokenArXiv54.6%unknown effort
All 1 recorded result & sources

54.6% · raw 54.6 %

Headline · unknown effort · Own vendor

Source/record date: Not recorded

from chart/figure Tencent Hy4-preview HF assets/benchmark-appendix.jpg; BrokenArXiv

https://huggingface.co/tencent/Hy4-preview

Independent evaluators

Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.

Independent evaluator headline scores; expand evidence for every record
Benchmark / evaluatorHeadline scoreEvidence
Vals Index · Vals AI49.9%unknown effort
All 1 recorded result & sources

49.9% · raw 49.94 %

Headline · unknown effort · Independent evaluator

Source/record date: 2026-10-03

Refreshed from current Vals leaderboard. Cost/test $1.44.

https://www.vals.ai/benchmarks/vals_index
Bugs fixed /105 · Bug Hunt Bench18 fixesunknown effort
All 1 recorded result & sources

18 fixes · raw 18 fixes

Headline · unknown effort · Independent evaluator

Source/record date: 2026-10-03

Harness: Claude Code / OpenRouter; effort default; 1 runs; evaluation 2026-08-28. Best documented score for this effort in Oct 1 README. Headline: best documented model run.

https://github.com/phuryn/bug-hunt-bench
Vibe Code Bench v1.1 · Vals AI77.5%unknown effort
All 1 recorded result & sources

77.5% · raw 77.48 %

Headline · unknown effort · Independent evaluator

Source/record date: 2026-10-03

Refreshed from current Vals leaderboard. Harness: OpenHands. Cost/test $2.18.

https://www.vals.ai/benchmarks/vibe-code

Read how we select and source scores or the comparison guide.