Skip to main content
BenchLM
Data

Furniture Assembly Benchmark (Epoch AI) (Furniture Assembly)

We show this table for reference; we do not rank on it.

Data verified

Epoch AI's Furniture Assembly benchmark shows models photos of IKEA furniture builds with the official manual and asks whether the build is correct so far and, if not, which step went wrong.

Benchmark score on Furniture Assembly — November 13, 2026

We compile the Furniture Assembly rows from benchmark-owner or independent runs. Claude Fable 5.2 leads the table at 90.0%, followed by Claude Opus 5.5 (83.3%) and GPT-6 Astra (80.0%). We do not use these results to rank models overall.

1Closed

Claude Fable 5.2

Anthropic

Run by Furniture Assembly
90.0%
Overall 93.65Context 1M
2Closed

Claude Opus 5.5

Anthropic

Run by Furniture Assembly
83.3%
Overall 86.62Context 1M
3Closed

GPT-6 Astra

OpenAI

Run by Furniture Assembly
80.0%
Overall 85.65Context 1.05M

37 modelsMultimodal & GroundedCurrentDisplay onlyUpdated November 13, 2026

Benchmark score table (37 models)

Benchmark score results for Furniture Assembly
RankModel / configurationScoreParameters (B)Open / closed
1Claude Fable 5.2AnthropicRun by Furniture Assembly
90.0%
Not reportedClosed
2Claude Opus 5.5AnthropicRun by Furniture Assembly
83.3%
Not reportedClosed
3GPT-6 AstraOpenAIRun by Furniture Assembly
80.0%
Not reportedClosed
4GPT-6.1 SolOpenAIRun by Furniture Assembly
80.0%
Not reportedClosed
5GPT-6.1 TerraOpenAIRun by Furniture Assembly
79.2%
Not reportedClosed
6Gemini 4 CarbonGoogleRun by Furniture Assembly
77.5%
Not reportedClosed
7Claude Sonnet 5.5AnthropicRun by Furniture Assembly
75.0%
Not reportedClosed
8Gemini 4 ArgonGoogleRun by Furniture Assembly
72.5%
Not reportedClosed
9Claude Fable 5.1AnthropicRun by Furniture Assembly
70.0%
Not reportedClosed
10Claude Opus 5AnthropicRun by Furniture Assembly
60.8%
Not reportedClosed
11GPT-6 SolOpenAIRun by Furniture Assembly
58.3%
Not reportedClosed
12GPT-5.6 SolOpenAIRun by Furniture Assembly
56.7%
Not reportedClosed
13Kimi K3.1 ProMoonshot AIRun by Furniture Assembly
55.8%
Not reportedOpen
14GPT-5.6 TerraOpenAIRun by Furniture Assembly
54.2%
Not reportedClosed
15Muse WonderMetaRun by Furniture Assembly
54.2%
Not reportedClosed
16Claude Haiku 5.5AnthropicRun by Furniture Assembly
47.5%
Not reportedClosed
17GPT-5.5OpenAIRun by Furniture Assembly
44.2%
Not reportedClosed
18GPT-6 LunaOpenAIRun by Furniture Assembly
44.2%
Not reportedClosed
19Claude Opus 4.8AnthropicRun by Furniture Assembly
42.5%
Not reportedClosed
20GPT-5.6 LunaOpenAIRun by Furniture Assembly
42.5%
Not reportedClosed
21Grok 4.6xAIRun by Furniture Assembly
40.0%
Not reportedClosed
22GPT-5.2OpenAIRun by Furniture Assembly
38.3%
Not reportedClosed
23GPT-5.4OpenAIRun by Furniture Assembly
37.5%
Not reportedClosed
24Claude Fable 5AnthropicRun by Furniture Assembly
35.8%
Not reportedClosed
25Kimi K3Moonshot AIRun by Furniture Assembly
34.2%
Not reportedOpen
26DeepSeek V4.1 Flash 1017DeepSeekRun by Furniture Assembly
34.2%
Not reportedOpen
27Claude Opus 4.7AnthropicRun by Furniture Assembly
33.3%
Not reportedClosed
28Gemini 3.8 FlashGoogleRun by Furniture Assembly
31.7%
Not reportedClosed
29Claude Opus 4.6AnthropicRun by Furniture Assembly
28.3%
Not reportedClosed
30Claude Opus 4.5AnthropicRun by Furniture Assembly
28.3%
Not reportedClosed
31Gemini 3.1 ProGoogleRun by Furniture Assembly
26.7%
Not reportedClosed
32Gemini 3.7 FlashGoogleRun by Furniture Assembly
26.7%
Not reportedClosed
33Gemini 3.6 FlashGoogleRun by Furniture Assembly
23.3%
Not reportedClosed
34Grok 4.5xAIRun by Furniture Assembly
22.5%
Not reportedClosed
35Kimi K2.6Moonshot AIRun by Furniture Assembly
21.7%
Not reportedOpen
36Grok 4.7xAIRun by Furniture Assembly
20.8%
Not reportedClosed
37Mistral Large 4MistralRun by Furniture Assembly
20.8%
Not reportedOpen
  • Lab-reported
  • Benchmark-run
  • Independent
What the labels mean

How to read this leaderboard

Higher is better. With 60 photos, one answer moves the score by about 1.7 points. Claude Fable 5.2 leads at 90%, ahead of Claude Opus 5.5 at 83.3%.

Operator receipt: 37 sourced rows are currently displayable on this page; the leading published row is Claude Fable 5.2 at 90.0%.

Honest limit: Three products and 60 photos make a small test set, and description grading relies on an LLM grader. It measures matching a manual to a real object, not physical manipulation. BenchLM shows it as display-only; it does not enter overall or category rankings.

Furniture Assembly leaderboard and methodologyEpoch AI benchmark data (CC-BY 4.0)

Among the reported Furniture Assembly rows, Claude Fable 5.2 is first at 90.0%. The third row is 10.0 points behind. The broader top-10 range is 29.2 points, so the table still separates the published systems.

37 models have been evaluated on Furniture Assembly. The benchmark falls in the Multimodal & Grounded category. We keep these results separate from the weighted global scoring system, so they remain source-specific evidence. Furniture Assembly is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Furniture Assembly

Year

2026

Tasks

60 photos of three IKEA builds, some with deliberate assembly mistakes

Format

Agentic visual check with manual PDF, zoom tool and Python, 80-step budget

Difficulty

Spatial reasoning from technical diagrams to real photos

Epoch built three IKEA products of increasing complexity (the Ställ shoe cabinet, Tonstad bed frame and Gullaberg dresser) and took 60 photos during assembly. Some show correct builds; others contain a deliberate, clearly visible mistake, often with assembly continued past it. For each photo the model gets the product, the step just completed, the photo, the assembly manual as a PDF, a zoom tool and a Python interpreter, with a budget of 80 agent steps. A correct build must be called correct; for a faulty build the model must name the mistaken step, and an LLM grader then checks the description leniently.

Furniture Assembly Benchmark (Epoch AI)

Freshness and provenance

Version

Furniture Assembly 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does the Furniture Assembly benchmark measure?

Spatial reasoning about physical construction: reading an IKEA manual, comparing it with a photo of a real build and spotting whether, and at which step, something was assembled wrong.

Which model is best at Furniture Assembly?

Claude Fable 5.2 at max effort leads with 90%, followed by Claude Opus 5.5 (83.3%), GPT-6.1 Sol and GPT-6 Astra (80% each).

Compare top models on Furniture Assembly

Claude Fable 5.2 vs Claude Opus 5.5Claude Opus 5.5 vs GPT-6 AstraGPT-6 Astra vs GPT-6.1 SolGPT-6.1 Sol vs GPT-6.1 Terra
Last updated: November 13, 2026 · BenchLM version Furniture Assembly 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.