benchgap
General

Furniture Assembly leaderboard

As of 2026-11-18, the highest measured score on Furniture Assembly is 90.0% by Claude Fable 5.2. 74 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.290.0%measured
2GPT-6.2 Astra89.2%measured
3GPT-6.2 Sol85.1%estimated ± 4.8 pp, medium confidence
4GPT-6 Cyber84.7%estimated ± 4.8 pp, medium confidence
5Claude Opus 5.583.3%measured
6GPT-6.1 Sol80.0%measured
7GPT-6 Astra80.0%measured
8GPT-6.1 Terra79.2%measured
9Gemini 4 Carbon77.5%measured
10Claude Mythos 5.277.4%estimated ± 6.5 pp, medium confidence
11Claude Sonnet 5.575.0%measured
12Gemini 4 Argon72.5%measured
13Claude Fable 5.170.0%measured
14Claude Mythos 566.3%estimated ± 6.5 pp, medium confidence
15Claude Opus 560.8%measured
16Claude Mythos Preview60.5%estimated ± 6.5 pp, medium confidence
17GPT-5.5 Pro59.7%estimated ± 14.5 pp, low confidence
18GPT-6 Sol58.3%measured
19GPT-5.6 Sol56.7%measured
20Kimi K3.1 Pro55.8%measured
21GPT-5.4 Pro55.5%estimated ± 14.5 pp, low confidence
22GPT-5.6 Terra54.2%measured
23Muse Wonder54.2%measured
24GLM-5.351.0%estimated ± 6.5 pp, medium confidence
25Claude Haiku 5.547.5%measured
26MiMo-V2.6-Pro46.8%estimated ± 6.5 pp, medium confidence
27Qwen3.8-27B46.5%estimated ± 12.4 pp, low confidence
28GPT-5.544.2%measured
29GPT-6 Luna44.2%measured
30Claude Opus 4.842.5%measured
31GPT-5.6 Luna42.5%measured
32Claude Sonnet 542.0%estimated ± 7.1 pp, medium confidence
33MiniMax M3.1 Flash40.1%estimated ± 12.4 pp, low confidence
34Grok 4.640.0%measured
35Muse Spark Open39.1%estimated ± 12.4 pp, low confidence
36Hy4 preview38.9%estimated ± 12.4 pp, low confidence
37SWE-2 Flash38.3%estimated ± 6.5 pp, medium confidence
38GPT-5.238.3%measured
39DeepSeek V4 Pro38.2%estimated ± 12.4 pp, low confidence
40Ling 3.1 Flash37.9%estimated ± 12.4 pp, low confidence
41GPT-5.437.5%measured
42Step 5 Preview36.2%estimated ± 12.4 pp, low confidence
43Qwen3.8 Max36.1%estimated ± 9.3 pp, medium confidence
44Claude Fable 535.8%measured
45DeepSeek V4 Flash 073134.5%estimated ± 9.3 pp, medium confidence
46DeepSeek V4.1 Flash 101734.2%measured
47Kimi K334.2%measured
48Beam 1.0 501B-A23B34.2%estimated ± 12.4 pp, low confidence
49Gemma 4 31B33.8%estimated ± 12.4 pp, low confidence
50o333.4%estimated ± 9.3 pp, medium confidence
51Gemma 4 26B A4B33.4%estimated ± 12.4 pp, low confidence
52Claude Opus 4.733.3%measured
53MAI-Thinking-133.3%estimated ± 12.4 pp, low confidence
54Muse Spark 1.333.0%estimated ± 9.3 pp, medium confidence
55Qwen3.6-35B-A3B32.9%estimated ± 9.3 pp, medium confidence
56Nemotron 3 Ultra32.8%estimated ± 9.3 pp, medium confidence
57Qwen3.5 Flash32.8%estimated ± 9.3 pp, medium confidence
58Qwen 3.6 Max (preview)32.8%estimated ± 9.3 pp, medium confidence
59Qwen3.5 397B32.8%estimated ± 9.3 pp, medium confidence
60Qwen3.5-122B-A10B32.8%estimated ± 9.3 pp, medium confidence
61Qwen3.5 Plus32.8%estimated ± 9.3 pp, medium confidence
62Qwen3.7 Plus32.8%estimated ± 9.3 pp, medium confidence
63Claude Sonnet 4.632.7%estimated ± 9.3 pp, medium confidence
64GPT-5.132.7%estimated ± 9.3 pp, medium confidence
65Qwen3.7 Flash32.7%estimated ± 9.3 pp, medium confidence
66GPT-4o mini32.7%estimated ± 9.3 pp, medium confidence
67Qwen3.6 Plus32.7%estimated ± 9.3 pp, medium confidence
68GLM-5.3-Flash32.7%estimated ± 9.3 pp, medium confidence
69MiniMax M332.7%estimated ± 9.3 pp, medium confidence
70GPT-4.1 mini32.7%estimated ± 9.3 pp, medium confidence
71GPT-5.4 mini32.7%estimated ± 9.3 pp, medium confidence
72o3-mini32.7%estimated ± 9.3 pp, medium confidence
73Qwen3.6-27B32.7%estimated ± 9.3 pp, medium confidence
74Inkling-Small32.7%estimated ± 9.3 pp, low confidence
75GPT-5.4 nano32.7%estimated ± 9.3 pp, low confidence
76o4-mini32.7%estimated ± 9.3 pp, low confidence
77Qwen3 Max32.7%estimated ± 9.3 pp, low confidence
78GPT-OSS 120B32.7%estimated ± 9.3 pp, low confidence
79Muse Spark 1.132.3%estimated ± 14.5 pp, low confidence
80MiMo-V2.6-Flash32.1%estimated ± 6.5 pp, low confidence
81Gemini 3.5 Flash-Lite32.1%estimated ± 7.1 pp, medium confidence
82Gemini 3 Flash32.1%estimated ± 7.1 pp, medium confidence
83Gemini 3.8 Flash31.7%measured
84Grok 4.2029.9%estimated ± 14.5 pp, low confidence
85Seed 2.1 Turbo28.8%estimated ± 14.5 pp, low confidence
86Claude Opus 4.528.3%measured
87Claude Opus 4.628.3%measured
88GPT-528.0%estimated ± 6.7 pp, medium confidence
89GLM-4.727.3%estimated ± 14.5 pp, low confidence
90Inkling27.3%estimated ± 14.5 pp, low confidence
91Seed 2.1 Pro27.3%estimated ± 14.5 pp, low confidence
92GLM-5.127.0%estimated ± 14.5 pp, low confidence
93Gemini 3.1 Pro26.7%measured
94Gemini 3.7 Flash26.7%measured
95Grok 4.326.5%estimated ± 14.5 pp, low confidence
96GLM-5.226.2%estimated ± 6.7 pp, medium confidence
97Qwen3.7 Max26.2%estimated ± 6.7 pp, medium confidence
98Kimi K2.7 Code25.7%estimated ± 14.5 pp, low confidence
99GPT-4.125.3%estimated ± 14.5 pp, low confidence
100o125.1%estimated ± 14.5 pp, low confidence
101GPT-4o25.0%estimated ± 14.5 pp, low confidence
102Claude Haiku 4.524.9%estimated ± 14.5 pp, low confidence
103Claude 3 Opus24.4%estimated ± 14.5 pp, low confidence
104GPT-4.1 nano24.1%estimated ± 14.5 pp, low confidence
105Gemini 3.5 Flash24.0%estimated ± 6.7 pp, medium confidence
106Muse Spark 1.223.4%estimated ± 12.1 pp, low confidence
107Gemini 3.6 Flash23.3%measured
108Claude Sonnet 4.523.0%estimated ± 6.7 pp, medium confidence
109Grok 4.522.5%measured
110Kimi K2.621.7%measured
111Grok 4.720.8%measured
112Mistral Large 420.8%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General