Skip to content
Weights Not Walls
Menu

Furniture Assembly

Re-run by Epoch AI

No 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.

Furniture Assembly results
#ModelWeightsSettingScoreDateSource
1claude-fable-5-2_maxAnthropicclosedeffort max90.0%±3.96 Nov 2026Epoch AI
2gpt-6.2-astra_maxOpenAIclosedeffort max89.2%±3.827 Oct 2026Epoch AI
3claude-opus-5-5_maxAnthropicclosedeffort max83.3%±4.722 Sep 2026Epoch AI
4gpt-6-astra_maxOpenAIclosedeffort max80.0%±4.910 Sep 2026Epoch AI
5gpt-6.1-sol_maxOpenAIclosedeffort max80.0%±4.929 Sep 2026Epoch AI
6gpt-6.1-terra_maxOpenAIclosedeffort max79.2%±5.126 Oct 2026Epoch AI
7gemini-4-carbon_highGoogle DeepMindclosedeffort high77.5%±5.43 Nov 2026Epoch AI
8claude-sonnet-5-5_maxAnthropicclosedeffort max75.0%±5.529 Sep 2026Epoch AI
9gemini-4-argon_highGoogle DeepMindclosedeffort high72.5%±5.613 Oct 2026Epoch AI
10claude-fable-5-1_maxAnthropicclosedeffort max70.0%±5.710 Sep 2026Epoch AI
11claude-opus-5_maxAnthropicclosedeffort max60.8%±6.210 Sep 2026Epoch AI
12gpt-6-sol_maxOpenAIclosedeffort max58.3%±6.423 Sep 2026Epoch AI
13gemini-4-hydrogen_xhighGoogle DeepMindclosedeffort xhigh57.5%±6.313 Nov 2026Epoch AI
14gpt-5.6-sol_maxOpenAIclosedeffort max56.7%±6.310 Sep 2026Epoch AI
15moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro-together_max, Moonshotopeneffort max55.8%±6.46 Nov 2026Epoch AI
16gpt-5.6-terra_maxOpenAIclosedeffort max54.2%±6.410 Sep 2026Epoch AI
17muse-wonder-1.0_maxMeta AIclosedeffort max54.2%±6.415 Oct 2026Epoch AI
18claude-haiku-5-5_maxAnthropicclosedeffort max47.5%±6.47 Oct 2026Epoch AI
19gpt-5.5_xhighOpenAIclosedeffort xhigh44.2%±6.410 Sep 2026Epoch AI
20gpt-6-luna_maxOpenAIclosedeffort max44.2%±6.428 Sep 2026Epoch AI
21claude-opus-4-8_maxAnthropicclosedeffort max42.5%±6.410 Sep 2026Epoch AI
22gpt-5.6-luna_maxOpenAIclosedeffort max42.5%±6.410 Sep 2026Epoch AI
23grok-4.6_xhighxAIclosedeffort xhigh40.0%±6.324 Sep 2026Epoch AI
24gpt-5.2-2025-12-11_xhighOpenAIclosedeffort xhigh38.3%±6.310 Sep 2026Epoch AI
25gpt-5.4-2026-03-05_xhighOpenAIclosedeffort xhigh37.5%±6.210 Sep 2026Epoch AI
26claude-fable-5_maxAnthropicclosedeffort max35.8%±6.210 Sep 2026Epoch AI
27deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_max, DeepSeekopeneffort max34.2%±6.08 Oct 2026Epoch AI
28moonshotai/Kimi-K3Epoch: kimi-k3_max, Moonshotopeneffort max34.2%±6.111 Sep 2026Epoch AI
29claude-opus-4-7_maxAnthropicclosedeffort max33.3%±6.110 Sep 2026Epoch AI
30gemini-3.8-flash_highGoogle DeepMindclosedeffort high31.7%±5.910 Sep 2026Epoch AI
31mistral-large-4-0-oss_highMistral AIno page matchopeneffort high30.8%±6.02 Nov 2026Epoch AI
32claude-opus-4-5-20251101_64KAnthropiccloseddefault28.3%±5.910 Sep 2026Epoch AI
33claude-opus-4-6_maxAnthropicclosedeffort max28.3%±5.910 Sep 2026Epoch AI
34gemini-3.1-pro-preview_highGoogle DeepMindclosedeffort high26.7%±5.610 Sep 2026Epoch AI
35gemini-3.7-flash_highGoogle DeepMindclosedeffort high26.7%±5.610 Sep 2026Epoch AI
36gemini-3.6-flash_highGoogle DeepMindclosedeffort high23.3%±5.410 Sep 2026Epoch AI
37grok-4.5_highxAIclosedeffort high22.5%±5.424 Sep 2026Epoch AI
38moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshotopendefault21.7%±5.211 Sep 2026Epoch AI
39grok-4.7_xhighxAIclosedeffort xhigh20.8%±5.224 Sep 2026Epoch AI
40mistral-large-4-0_highMistral AIno page matchopeneffort high20.8%±5.214 Oct 2026Epoch AI
41qwen3.8-max-0902_xhighAlibabaclosedeffort xhigh20.0%±5.210 Sep 2026Epoch 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