GSO-Bench
ReportedNo description yet: we only describe an eval set when we can do it correctly.
- Kind
- Published by a third party's leaderboard, collected by Epoch AI. The leaderboard is named and linked.
- Link
- https://gso-bench.github.io/index.html
- Score shown
- Epoch's column "Score OPT@1", as a percentage, higher is better; random guessing scores about 0.0%. The source gives no error bars.
- Results
- 42 in all: 3 for open-weight models, 39 for closed ones. 3 linked to a model page here.
Best open-weight result: moonshotai/Kimi-K2-Instruct, 4.9%, number 31 of 42 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | claude-fable-5-1_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 88.2% | 27 Sep 2026 | gso-bench.github.io |
| 2 | gpt-6-astra_unknownOpenAI | closed | effort unknown; Scaffold: OpenHands | 79.4% | 27 Sep 2026 | gso-bench.github.io |
| 3 | claude-fable-5_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 78.4% | 27 Sep 2026 | gso-bench.github.io |
| 4 | gpt-5.6-sol_unknownOpenAI | closed | effort unknown; Scaffold: OpenHands | 76.5% | 27 Sep 2026 | gso-bench.github.io |
| 5 | claude-opus-4-8_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 47.1% | 12 Jul 2026 | gso-bench.github.io |
| 6 | claude-opus-4-7Anthropic | closed | Scaffold: OpenHands | 44.1% | 27 Apr 2026 | GSO Leaderboard, via Epoch AI |
| 7 | claude-opus-4-7_highAnthropic | closed | effort high; Scaffold: OpenHands | 44.1% | 27 Apr 2026 | gso-bench.github.io |
| 8 | claude-opus-4-6_highAnthropic | closed | effort high; Scaffold: OpenHands | 41.2% | 27 Apr 2026 | GSO Leaderboard, via Epoch AI |
| 9 | gpt-5.5_xhighOpenAI | closed | effort xhigh; Scaffold: OpenHands | 40.2% | 27 Apr 2026 | gso-bench.github.io |
| 10 | claude-sonnet-5_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 37.3% | 12 Jul 2026 | gso-bench.github.io |
| 11 | claude-opus-4-6Anthropic | closed | Scaffold: OpenHands | 33.3% | 27 Apr 2026 | GSO Leaderboard, via Epoch AI |
| 12 | claude-opus-4-6_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 33.3% | 27 Apr 2026 | gso-bench.github.io |
| 13 | gpt-5.4-2026-03-05_xhighOpenAI | closed | effort xhigh; Scaffold: OpenHands | 31.4% | 10 Mar 2026 | gso-bench.github.io |
| 14 | gpt-5.2-2025-12-11_highOpenAI | closed | effort high; Scaffold: OpenHands | 27.5% | 18 Dec 2025 | gso-bench.github.io |
| 15 | claude-opus-4-5-20251101Anthropic | closed | Scaffold: OpenHands | 26.5% | not dated | GSO Leaderboard |
| 16 | claude-opus-4-5-20251101_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 26.5% | 28 Nov 2025 | gso-bench.github.io |
| 17 | gpt-5.4-2026-03-05_highOpenAI | closed | effort high; Scaffold: OpenHands | 25.5% | 10 Mar 2026 | gso-bench.github.io |
| 18 | gemini-3.1-pro-previewGoogle DeepMind | closed | Scaffold: OpenHands | 22.6% | 9 Mar 2026 | gso-bench.github.io |
| 19 | gemini-3-pro-previewGoogle DeepMind | closed | Scaffold: OpenHands | 18.6% | 28 Nov 2025 | gso-bench.github.io |
| 20 | claude-sonnet-4-5-20250929_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 14.7% | 10 Oct 2025 | gso-bench.github.io |
| 21 | claude-sonnet-4-5-20250929Anthropic | closed | Scaffold: OpenHands | 14.7% | not dated | GSO Leaderboard |
| 22 | gpt-5.1-2025-11-13_unknownOpenAI | closed | effort unknown; Scaffold: OpenHands | 13.7% | 18 Dec 2025 | gso-bench.github.io |
| 23 | gpt-5.1-2025-11-13_highOpenAI | closed | effort high; Scaffold: OpenHands | 13.7% | not dated | GSO Leaderboard, via Epoch AI |
| 24 | gemini-3-flash-previewGoogle DeepMind | closed | Scaffold: OpenHands | 9.8% | 19 Dec 2025 | gso-bench.github.io |
| 25 | o3-2025-04-16_highOpenAI | closed | effort high; Scaffold: OpenHands | 8.8% | 9 Jun 2025 | gso-bench.github.io |
| 26 | claude-opus-4-20250514Anthropic | closed | Scaffold: OpenHands | 6.9% | not dated | GSO Leaderboard |
| 27 | claude-opus-4-20250514_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 6.9% | 10 Aug 2025 | gso-bench.github.io |
| 28 | gpt-5-2025-08-07_highOpenAI | closed | effort high; Scaffold: OpenHands | 6.9% | 10 Aug 2025 | gso-bench.github.io |
| 29 | claude-sonnet-4-20250514Anthropic | closed | Scaffold: OpenHands | 4.9% | not dated | GSO Leaderboard |
| 30 | claude-sonnet-4-20250514_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 4.9% | 30 May 2025 | gso-bench.github.io |
| 31 | moonshotai/Kimi-K2-InstructEpoch: Kimi-K2-Instruct, Moonshot | open | Scaffold: OpenHands | 4.9% | 26 Aug 2025 | gso-bench.github.io |
| 32 | Qwen/Qwen3-Coder-480B-A35B-InstructEpoch: Qwen3-Coder-480B-A35B-Instruct, Alibaba | open | Scaffold: OpenHands | 4.9% | 2 Sep 2025 | gso-bench.github.io |
| 33 | claude-3-5-sonnet-20241022Anthropic | closed | Scaffold: OpenHands | 4.6% | 30 May 2025 | gso-bench.github.io |
| 34 | gemini-2.5-proGoogle DeepMind | closed | Scaffold: OpenHands | 3.9% | 10 Oct 2025 | gso-bench.github.io |
| 35 | gemini-2.5-pro-preview-06-05Google DeepMind | closed | Scaffold: OpenHands | 3.9% | 10 Oct 2025 | GSO Leaderboard, via Epoch AI |
| 36 | claude-3-7-sonnet-20250219Anthropic | closed | Scaffold: OpenHands | 3.8% | not dated | GSO Leaderboard |
| 37 | claude-3-7-sonnet-20250219_unknownAnthropic | closed | effort unknown; Scaffold: OpenHands | 3.8% | 30 May 2025 | gso-bench.github.io |
| 38 | o4-mini-2025-04-16_highOpenAI | closed | effort high; Scaffold: OpenHands | 3.6% | 30 May 2025 | gso-bench.github.io |
| 39 | zai-org/GLM-4.5-AirEpoch: GLM-4.5-Air, Z.ai (Zhipu AI),Tsinghua University | open | Scaffold: OpenHands | 2.9% | 26 Aug 2025 | gso-bench.github.io |
| 40 | o3-mini-2025-01-31_highOpenAI | closed | effort high; Scaffold: OpenHands | 1.3% | 30 May 2025 | gso-bench.github.io |
| 41 | o3-mini-2025-01-31_lowOpenAI | closed | effort low; Scaffold: OpenHands | 1.3% | 30 May 2025 | GSO Leaderboard, via Epoch AI |
| 42 | gpt-4o-2024-11-20OpenAI | closed | Scaffold: OpenHands | 0.0% | 30 May 2025 | gso-bench.github.io |
Open-weight rows are shaded. Dates are when the result was run or evaluated, where the data says. Data from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Epoch AI. Method