Furniture Assembly
Re-run by Epoch AINo description yet: we only describe an eval set when we can do it correctly.
- Kind
- Run independently by Epoch AI, with public logs. Our own re-runs are not built yet.
- Score shown
- Epoch's column "Best score (across scorers)", as a percentage, higher is better; random guessing scores about 30.0%. Error bars as the source gives them.
- Results
- 41 in all: 6 for open-weight models, 35 for closed ones. 4 linked to a model page here.
Best open-weight result: moonshotai/Kimi-K3.1-Pro, 55.8%, number 15 of 41 results.
| # | Model | Weights | Setting | Score | Date | Source |
|---|---|---|---|---|---|---|
| 1 | claude-fable-5-2_maxAnthropic | closed | effort max | 90.0%±3.9 | 6 Nov 2026 | Epoch AI |
| 2 | gpt-6.2-astra_maxOpenAI | closed | effort max | 89.2%±3.8 | 27 Oct 2026 | Epoch AI |
| 3 | claude-opus-5-5_maxAnthropic | closed | effort max | 83.3%±4.7 | 22 Sep 2026 | Epoch AI |
| 4 | gpt-6-astra_maxOpenAI | closed | effort max | 80.0%±4.9 | 10 Sep 2026 | Epoch AI |
| 5 | gpt-6.1-sol_maxOpenAI | closed | effort max | 80.0%±4.9 | 29 Sep 2026 | Epoch AI |
| 6 | gpt-6.1-terra_maxOpenAI | closed | effort max | 79.2%±5.1 | 26 Oct 2026 | Epoch AI |
| 7 | gemini-4-carbon_highGoogle DeepMind | closed | effort high | 77.5%±5.4 | 3 Nov 2026 | Epoch AI |
| 8 | claude-sonnet-5-5_maxAnthropic | closed | effort max | 75.0%±5.5 | 29 Sep 2026 | Epoch AI |
| 9 | gemini-4-argon_highGoogle DeepMind | closed | effort high | 72.5%±5.6 | 13 Oct 2026 | Epoch AI |
| 10 | claude-fable-5-1_maxAnthropic | closed | effort max | 70.0%±5.7 | 10 Sep 2026 | Epoch AI |
| 11 | claude-opus-5_maxAnthropic | closed | effort max | 60.8%±6.2 | 10 Sep 2026 | Epoch AI |
| 12 | gpt-6-sol_maxOpenAI | closed | effort max | 58.3%±6.4 | 23 Sep 2026 | Epoch AI |
| 13 | gemini-4-hydrogen_xhighGoogle DeepMind | closed | effort xhigh | 57.5%±6.3 | 13 Nov 2026 | Epoch AI |
| 14 | gpt-5.6-sol_maxOpenAI | closed | effort max | 56.7%±6.3 | 10 Sep 2026 | Epoch AI |
| 15 | moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro-together_max, Moonshot | open | effort max | 55.8%±6.4 | 6 Nov 2026 | Epoch AI |
| 16 | gpt-5.6-terra_maxOpenAI | closed | effort max | 54.2%±6.4 | 10 Sep 2026 | Epoch AI |
| 17 | muse-wonder-1.0_maxMeta AI | closed | effort max | 54.2%±6.4 | 15 Oct 2026 | Epoch AI |
| 18 | claude-haiku-5-5_maxAnthropic | closed | effort max | 47.5%±6.4 | 7 Oct 2026 | Epoch AI |
| 19 | gpt-5.5_xhighOpenAI | closed | effort xhigh | 44.2%±6.4 | 10 Sep 2026 | Epoch AI |
| 20 | gpt-6-luna_maxOpenAI | closed | effort max | 44.2%±6.4 | 28 Sep 2026 | Epoch AI |
| 21 | claude-opus-4-8_maxAnthropic | closed | effort max | 42.5%±6.4 | 10 Sep 2026 | Epoch AI |
| 22 | gpt-5.6-luna_maxOpenAI | closed | effort max | 42.5%±6.4 | 10 Sep 2026 | Epoch AI |
| 23 | grok-4.6_xhighxAI | closed | effort xhigh | 40.0%±6.3 | 24 Sep 2026 | Epoch AI |
| 24 | gpt-5.2-2025-12-11_xhighOpenAI | closed | effort xhigh | 38.3%±6.3 | 10 Sep 2026 | Epoch AI |
| 25 | gpt-5.4-2026-03-05_xhighOpenAI | closed | effort xhigh | 37.5%±6.2 | 10 Sep 2026 | Epoch AI |
| 26 | claude-fable-5_maxAnthropic | closed | effort max | 35.8%±6.2 | 10 Sep 2026 | Epoch AI |
| 27 | deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_max, DeepSeek | open | effort max | 34.2%±6.0 | 8 Oct 2026 | Epoch AI |
| 28 | moonshotai/Kimi-K3Epoch: kimi-k3_max, Moonshot | open | effort max | 34.2%±6.1 | 11 Sep 2026 | Epoch AI |
| 29 | claude-opus-4-7_maxAnthropic | closed | effort max | 33.3%±6.1 | 10 Sep 2026 | Epoch AI |
| 30 | gemini-3.8-flash_highGoogle DeepMind | closed | effort high | 31.7%±5.9 | 10 Sep 2026 | Epoch AI |
| 31 | mistral-large-4-0-oss_highMistral AIno page match | open | effort high | 30.8%±6.0 | 2 Nov 2026 | Epoch AI |
| 32 | claude-opus-4-5-20251101_64KAnthropic | closed | default | 28.3%±5.9 | 10 Sep 2026 | Epoch AI |
| 33 | claude-opus-4-6_maxAnthropic | closed | effort max | 28.3%±5.9 | 10 Sep 2026 | Epoch AI |
| 34 | gemini-3.1-pro-preview_highGoogle DeepMind | closed | effort high | 26.7%±5.6 | 10 Sep 2026 | Epoch AI |
| 35 | gemini-3.7-flash_highGoogle DeepMind | closed | effort high | 26.7%±5.6 | 10 Sep 2026 | Epoch AI |
| 36 | gemini-3.6-flash_highGoogle DeepMind | closed | effort high | 23.3%±5.4 | 10 Sep 2026 | Epoch AI |
| 37 | grok-4.5_highxAI | closed | effort high | 22.5%±5.4 | 24 Sep 2026 | Epoch AI |
| 38 | moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshot | open | default | 21.7%±5.2 | 11 Sep 2026 | Epoch AI |
| 39 | grok-4.7_xhighxAI | closed | effort xhigh | 20.8%±5.2 | 24 Sep 2026 | Epoch AI |
| 40 | mistral-large-4-0_highMistral AIno page match | open | effort high | 20.8%±5.2 | 14 Oct 2026 | Epoch AI |
| 41 | qwen3.8-max-0902_xhighAlibaba | closed | effort xhigh | 20.0%±5.2 | 10 Sep 2026 | Epoch AI |
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