Inference
EN
Epoch AI

Furniture Assembly

Evaluated by Epoch AI on their own harness.

ModelsModels with a result.37
Top resultBest result on this benchmark.90.0%Claude Fable 5.2
Top-3 spreadPoints from first to third.10.0 pts
YearYear of release.—

Result and price

37 of 37 results

Results

37 results
#ModelResultThe score from the source.
01Claude Fable 5.2max effort90.0% ±3.9
02Claude Opus 5.5max effort83.3% ±4.7
03GPT-6 Astramax effort80.0% ±4.9
03GPT-6.1 Solmax effort80.0% ±4.9
05GPT-6.1 Terramax effort79.2% ±5.1
06Gemini 4 Carbonhigh effort77.5% ±5.4
07Claude Sonnet 5.5max effort75.0% ±5.5
08Gemini 4 Argonhigh effort72.5% ±5.6
09Claude Fable 5.1max effort70.0% ±5.7
10Claude Opus 5max effort60.8% ±6.2
11GPT-6 Solmax effort58.3% ±6.4
12GPT-5.6 Solmax effort56.7% ±6.3
13Muse Wondermax effort54.2% ±6.4
13GPT-5.6 Terramax effort54.2% ±6.4
15Claude Haiku 5.5max effort47.5% ±6.4
16GPT-5.5xhigh effort44.2% ±6.4
16GPT-6 Lunamax effort44.2% ±6.4
18Claude Opus 4.8max effort42.5% ±6.4
18GPT-5.6 Lunamax effort42.5% ±6.4
20Grok 4.6xhigh effort40.0% ±6.3
21GPT-5.2xhigh effort38.3% ±6.3
22GPT-5.4xhigh effort37.5% ±6.2
23Claude Fable 5max effort35.8% ±6.2
24DeepSeek V4.1 Flash 1017max effort34.2% ±6.0
24Kimi K3max effort34.2% ±6.1
26Claude Opus 4.7max effort33.3% ±6.1
27Gemini 3.8 Flashhigh effort31.7% ±5.9
28Claude Opus 4.6max effort28.3% ±5.9
28Claude Opus 4.528.3% ±5.9
30Gemini 3.7 Flashhigh effort26.7% ±5.6
30Gemini 3.1 Prohigh effort26.7% ±5.6
32Mistral Large 4high effort25.8% ±6.0
33Gemini 3.6 Flashhigh effort23.3% ±5.4
34Grok 4.5high effort22.5% ±5.4
35Kimi K2.621.7% ±5.2
36Grok 4.7xhigh effort20.8% ±5.2
37Qwen3.8 Max (0902)xhigh effort20.0% ±5.2