Benchmark Almanac

Furniture Assembly Benchmark

Claude Fable 5.2 90% leads since November 2026
Claude Fable 5.2 90% leads since November 2026
Not in model cards

Model gets an IKEA manual and a photo of a partly built item, then names the faulty step or next step. Accuracy on 60 photos.

LLM grader checks error descriptions. Novice human baseline 63%. Photos and manuals are not distributed.

Made by
Epoch AI
Released
September 2026
Kind
Spatial reasoning and counting
Level
Everyday
Scored
Graded by a model
Answers
Short answer
Tasks
Kept private
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last model added
November 2026
Scores refreshed
November 12th, 2026
39
Entries on its own leaderboard, 35 models we track

Claude Fable 5.2 leads with 90%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

Claude Fable 5.2 leads with 90% across all independent runs

Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.

39 entries. Scroll sideways for the rest.

  1. 90.0 Claude Fable 5.2
  2. 83.3 Claude Opus 5.5
  3. 80.0 GPT-6 Astra
  4. 80.0 GPT-6.1 Sol
  5. 79.2 GPT-6.1 Terra
  6. 77.5 Gemini 4 Carbon
  7. 75.0 Claude Sonnet 5.5
  8. 72.5 Gemini 4 Argon
  9. 70.0 Claude Fable 5.1
  10. 60.8 Claude Opus 5
  11. 58.3 GPT-6 Sol
  12. 56.7 GPT-5.6 Sol
  13. 55.8 kimi-k3.1-pro-together
  14. 54.2 GPT-5.6 Terra
  15. 54.2 muse-wonder-1.0
  16. 47.5 Claude Haiku 5.5
  17. 44.2 GPT-5.5
  18. 44.2 GPT-6 Luna
  19. 42.5 Claude Opus 4.8
  20. 42.5 GPT-5.6 Luna
  21. 40.0 Grok 4.6
  22. 38.3 GPT-5.2
  23. 37.5 GPT-5.4
  24. 35.8 Claude Fable 5
  25. 34.2 Kimi K3
  26. 34.2 DeepSeek V4.1 Flash
  27. 33.3 Claude Opus 4.7
  28. 31.7 Gemini 3.8 Flash
  29. 30.8 mistral-large-4-0-oss
  30. 28.3 Claude Opus 4.5
  31. 28.3 Claude Opus 4.6
  32. 26.7 Gemini 3.1 Pro
  33. 26.7 Gemini 3.7 Flash
  34. 23.3 Gemini 3.6 Flash
  35. 22.5 Grok 4.5
  36. 21.7 Kimi K2.6
  37. 20.8 Grok 4.7
  38. 20.8 mistral-large-4-0
  39. 20.0 Qwen3.8 Max (0902)
  1. 1 Claude Fable 5.2 max; task version 1.0.1 90%
  2. 2 Claude Opus 5.5 max; task version 1.0.0 83.3%
  3. 3 GPT-6 Astra max; task version 1.0.0 80%
  4. 4 GPT-6.1 Sol max; task version 1.0.0 80%
  5. 5 GPT-6.1 Terra max; task version 1.0.1 79.2%
  6. 6 Gemini 4 Carbon high; task version 1.0.1 77.5%
  7. 7 Claude Sonnet 5.5 max; task version 1.0.0 75%
  8. 8 Gemini 4 Argon high; task version 1.0.1 72.5%
  9. 9 Claude Fable 5.1 max; task version 1.0.0 70%
  10. 10 Claude Opus 5 max; task version 1.0.0 60.8%
  11. 11 GPT-6 Sol max; task version 1.0.0 58.3%
  12. 12 GPT-5.6 Sol max; task version 1.0.0 56.7%
  13. 13 kimi-k3.1-pro-together max; task version 1.0.1 55.8%
  14. 14 GPT-5.6 Terra max; task version 1.0.0 54.2%
  15. 15 muse-wonder-1.0 max; task version 1.0.1 54.2%
  16. 16 Claude Haiku 5.5 max; task version 1.0.0 47.5%
  17. 17 GPT-5.5 xhigh; task version 1.0.0 44.2%
  18. 18 GPT-6 Luna max; task version 1.0.0 44.2%
  19. 19 Claude Opus 4.8 max; task version 1.0.0 42.5%
  20. 20 GPT-5.6 Luna max; task version 1.0.0 42.5%
  21. 21 Grok 4.6 xhigh; task version 1.0.0 40%
  22. 22 GPT-5.2 xhigh; task version 1.0.0 38.3%
  23. 23 GPT-5.4 xhigh; task version 1.0.0 37.5%
  24. 24 Claude Fable 5 max; task version 1.0.0 35.8%
  25. 25 Kimi K3 max; task version 1.0.0 34.2%
  26. 26 DeepSeek V4.1 Flash max; task version 1.0.0 34.2%
  27. 27 Claude Opus 4.7 max; task version 1.0.0 33.3%
  28. 28 Gemini 3.8 Flash high; task version 1.0.0 31.7%
  29. 29 mistral-large-4-0-oss high; task version 1.0.1 30.8%
  30. 30 Claude Opus 4.5 64K; task version 1.0.0 28.3%
  31. 31 Claude Opus 4.6 max; task version 1.0.0 28.3%
  32. 32 Gemini 3.1 Pro high; task version 1.0.0 26.7%
  33. 33 Gemini 3.7 Flash high; task version 1.0.0 26.7%
  34. 34 Gemini 3.6 Flash high; task version 1.0.0 23.3%
  35. 35 Grok 4.5 high; task version 1.0.0 22.5%
  36. 36 Kimi K2.6 task version 1.0.0 21.7%
  37. 37 Grok 4.7 xhigh; task version 1.0.0 20.8%
  38. 38 mistral-large-4-0 high; task version 1.0.1 20.8%
  39. 39 Qwen3.8 Max (0902) xhigh; task version 1.0.0 20%

Source: Official Furniture Assembly Benchmark leaderboard

39 models. Scroll sideways for the rest.

  1. 90.0 Claude Fable 5.2
  2. 83.3 Claude Opus 5.5
  3. 80.0 GPT-6 Astra
  4. 80.0 GPT-6.1 Sol
  5. 79.2 GPT-6.1 Terra
  6. 77.5 Gemini 4 Carbon
  7. 75.0 Claude Sonnet 5.5
  8. 72.5 Gemini 4 Argon
  9. 70.0 Claude Fable 5.1
  10. 60.8 Claude Opus 5
  11. 58.3 GPT-6 Sol
  12. 56.7 GPT-5.6 Sol
  13. 55.8 Kimi K3.1 Pro
  14. 54.2 GPT-5.6 Terra
  15. 54.2 Muse Wonder
  16. 47.5 Claude Haiku 5.5
  17. 44.2 GPT-5.5
  18. 44.2 GPT-6 Luna
  19. 42.5 Claude Opus 4.8
  20. 42.5 GPT-5.6 Luna
  21. 40.0 Grok 4.6
  22. 38.3 GPT-5.2
  23. 37.5 GPT-5.4
  24. 35.8 Claude Fable 5
  25. 34.2 Kimi K3
  26. 34.2 DeepSeek V4.1 Flash
  27. 33.3 Claude Opus 4.7
  28. 31.7 Gemini 3.8 Flash
  29. 30.8 Mistral Large 4
  30. 28.3 Claude Opus 4.5
  31. 28.3 Claude Opus 4.6
  32. 26.7 Gemini 3.1 Pro
  33. 26.7 Gemini 3.7 Flash
  34. 23.3 Gemini 3.6 Flash
  35. 22.5 Grok 4.5
  36. 21.7 Kimi K2.6
  37. 20.8 Grok 4.7
  38. 20.8 Mistral Large 4 Preview
  39. 20.0 Qwen3.8 Max (0902)
  1. 1 Claude Fable 5.2 max; task version 1.0.1 90%
  2. 2 Claude Opus 5.5 max; task version 1.0.0 83.3%
  3. 3 GPT-6 Astra max; task version 1.0.0 80%
  4. 4 GPT-6.1 Sol max; task version 1.0.0 80%
  5. 5 GPT-6.1 Terra max; task version 1.0.1 79.2%
  6. 6 Gemini 4 Carbon high; task version 1.0.1 77.5%
  7. 7 Claude Sonnet 5.5 max; task version 1.0.0 75%
  8. 8 Gemini 4 Argon high; task version 1.0.1 72.5%
  9. 9 Claude Fable 5.1 max; task version 1.0.0 70%
  10. 10 Claude Opus 5 max; task version 1.0.0 60.8%
  11. 11 GPT-6 Sol max; task version 1.0.0 58.3%
  12. 12 GPT-5.6 Sol max; task version 1.0.0 56.7%
  13. 13 Kimi K3.1 Pro Together, max 55.8%
  14. 14 GPT-5.6 Terra max; task version 1.0.0 54.2%
  15. 15 Muse Wonder max 54.2%
  16. 16 Claude Haiku 5.5 max; task version 1.0.0 47.5%
  17. 17 GPT-5.5 xhigh; task version 1.0.0 44.2%
  18. 18 GPT-6 Luna max; task version 1.0.0 44.2%
  19. 19 Claude Opus 4.8 max; task version 1.0.0 42.5%
  20. 20 GPT-5.6 Luna max; task version 1.0.0 42.5%
  21. 21 Grok 4.6 xhigh; task version 1.0.0 40%
  22. 22 GPT-5.2 xhigh; task version 1.0.0 38.3%
  23. 23 GPT-5.4 xhigh; task version 1.0.0 37.5%
  24. 24 Claude Fable 5 max; task version 1.0.0 35.8%
  25. 25 Kimi K3 max; task version 1.0.0 34.2%
  26. 26 DeepSeek V4.1 Flash max; task version 1.0.0 34.2%
  27. 27 Claude Opus 4.7 max; task version 1.0.0 33.3%
  28. 28 Gemini 3.8 Flash high; task version 1.0.0 31.7%
  29. 29 Mistral Large 4 high 30.8%
  30. 30 Claude Opus 4.5 64K; task version 1.0.0 28.3%
  31. 31 Claude Opus 4.6 max; task version 1.0.0 28.3%
  32. 32 Gemini 3.1 Pro high; task version 1.0.0 26.7%
  33. 33 Gemini 3.7 Flash high; task version 1.0.0 26.7%
  34. 34 Gemini 3.6 Flash high; task version 1.0.0 23.3%
  35. 35 Grok 4.5 high; task version 1.0.0 22.5%
  36. 36 Kimi K2.6 task version 1.0.0 21.7%
  37. 37 Grok 4.7 xhigh; task version 1.0.0 20.8%
  38. 38 Mistral Large 4 Preview high 20.8%
  39. 39 Qwen3.8 Max (0902) xhigh; task version 1.0.0 20%

Sources: Official Furniture Assembly Benchmark leaderboard and Epoch AI runs (CC BY 4.0)

#EntryScoreSettingDateSource
1 Claude Fable 5.2 90% max; task version 1.0.1, Epoch Nov 2026 Official leaderboard
2 Claude Opus 5.5 83.3% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
3 GPT-6 Astra 80% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
4 GPT-6.1 Sol 80% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
5 GPT-6.1 Terra 79.2% max; task version 1.0.1, Epoch Oct 2026 Official leaderboard
6 Gemini 4 Carbon 77.5% high; task version 1.0.1, Epoch Oct 2026 Official leaderboard
7 Claude Sonnet 5.5 75% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
8 Gemini 4 Argon 72.5% high; task version 1.0.1, Epoch Sep 2026 Official leaderboard
9 Claude Fable 5.1 70% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
10 Claude Opus 5 60.8% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
11 GPT-6 Sol 58.3% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
12 GPT-5.6 Sol 56.7% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
13 kimi-k3.1-pro-together 55.8% max; task version 1.0.1, Epoch Nov 2026 Official leaderboard
14 GPT-5.6 Terra 54.2% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
15 muse-wonder-1.0 54.2% max; task version 1.0.1, Epoch Oct 2026 Official leaderboard
16 Claude Haiku 5.5 47.5% max; task version 1.0.0, Epoch Oct 2026 Official leaderboard
17 GPT-5.5 44.2% xhigh; task version 1.0.0, Epoch Apr 2026 Official leaderboard
18 GPT-6 Luna 44.2% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
19 Claude Opus 4.8 42.5% max; task version 1.0.0, Epoch May 2026 Official leaderboard
20 GPT-5.6 Luna 42.5% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
21 Grok 4.6 40% xhigh; task version 1.0.0, Epoch Aug 2026 Official leaderboard
22 GPT-5.2 38.3% xhigh; task version 1.0.0, Epoch Dec 2025 Official leaderboard
23 GPT-5.4 37.5% xhigh; task version 1.0.0, Epoch Mar 2026 Official leaderboard
24 Claude Fable 5 35.8% max; task version 1.0.0, Epoch Jun 2026 Official leaderboard
25 Kimi K3 34.2% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
26 DeepSeek V4.1 Flash 34.2% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
27 Claude Opus 4.7 33.3% max; task version 1.0.0, Epoch Apr 2026 Official leaderboard
28 Gemini 3.8 Flash 31.7% high; task version 1.0.0, Epoch Sep 2026 Official leaderboard
29 mistral-large-4-0-oss 30.8% high; task version 1.0.1, Epoch Nov 2026 Official leaderboard
30 Claude Opus 4.5 28.3% 64K; task version 1.0.0, Epoch Nov 2025 Official leaderboard
31 Claude Opus 4.6 28.3% max; task version 1.0.0, Epoch Feb 2026 Official leaderboard
32 Gemini 3.1 Pro 26.7% high; task version 1.0.0, Epoch Feb 2026 Official leaderboard
33 Gemini 3.7 Flash 26.7% high; task version 1.0.0, Epoch Aug 2026 Official leaderboard
34 Gemini 3.6 Flash 23.3% high; task version 1.0.0, Epoch Jul 2026 Official leaderboard
35 Grok 4.5 22.5% high; task version 1.0.0, Epoch Jul 2026 Official leaderboard
36 Kimi K2.6 21.7% task version 1.0.0, Epoch Apr 2026 Official leaderboard
37 Grok 4.7 20.8% xhigh; task version 1.0.0, Epoch Sep 2026 Official leaderboard
38 mistral-large-4-0 20.8% high; task version 1.0.1, Epoch Oct 2026 Official leaderboard
39 Qwen3.8 Max (0902) 20% xhigh; task version 1.0.0, Epoch Sep 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Fable 5.2 90% max; task version 1.0.1, Epoch Nov 2026 Official leaderboard
2 Claude Opus 5.5 83.3% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
3 GPT-6 Astra 80% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
4 GPT-6.1 Sol 80% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
5 GPT-6.1 Terra 79.2% max; task version 1.0.1, Epoch Oct 2026 Official leaderboard
6 Gemini 4 Carbon 77.5% high; task version 1.0.1, Epoch Oct 2026 Official leaderboard
7 Claude Sonnet 5.5 75% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
8 Gemini 4 Argon 72.5% high; task version 1.0.1, Epoch Sep 2026 Official leaderboard
9 Claude Fable 5.1 70% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
10 Claude Opus 5 60.8% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
11 GPT-6 Sol 58.3% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
12 GPT-5.6 Sol 56.7% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
13 Kimi K3.1 Pro 55.8% Together, max Oct 2026 Epoch AI run
14 GPT-5.6 Terra 54.2% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
15 Muse Wonder 54.2% max Oct 2026 Epoch AI run
16 Claude Haiku 5.5 47.5% max; task version 1.0.0, Epoch Oct 2026 Official leaderboard
17 GPT-5.5 44.2% xhigh; task version 1.0.0, Epoch Apr 2026 Official leaderboard
18 GPT-6 Luna 44.2% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
19 Claude Opus 4.8 42.5% max; task version 1.0.0, Epoch May 2026 Official leaderboard
20 GPT-5.6 Luna 42.5% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
21 Grok 4.6 40% xhigh; task version 1.0.0, Epoch Aug 2026 Official leaderboard
22 GPT-5.2 38.3% xhigh; task version 1.0.0, Epoch Dec 2025 Official leaderboard
23 GPT-5.4 37.5% xhigh; task version 1.0.0, Epoch Mar 2026 Official leaderboard
24 Claude Fable 5 35.8% max; task version 1.0.0, Epoch Jun 2026 Official leaderboard
25 Kimi K3 34.2% max; task version 1.0.0, Epoch Jul 2026 Official leaderboard
26 DeepSeek V4.1 Flash 34.2% max; task version 1.0.0, Epoch Sep 2026 Official leaderboard
27 Claude Opus 4.7 33.3% max; task version 1.0.0, Epoch Apr 2026 Official leaderboard
28 Gemini 3.8 Flash 31.7% high; task version 1.0.0, Epoch Sep 2026 Official leaderboard
29 Mistral Large 4 30.8% high Oct 2026 Epoch AI run
30 Claude Opus 4.5 28.3% 64K; task version 1.0.0, Epoch Nov 2025 Official leaderboard
31 Claude Opus 4.6 28.3% max; task version 1.0.0, Epoch Feb 2026 Official leaderboard
32 Gemini 3.1 Pro 26.7% high; task version 1.0.0, Epoch Feb 2026 Official leaderboard
33 Gemini 3.7 Flash 26.7% high; task version 1.0.0, Epoch Aug 2026 Official leaderboard
34 Gemini 3.6 Flash 23.3% high; task version 1.0.0, Epoch Jul 2026 Official leaderboard
35 Grok 4.5 22.5% high; task version 1.0.0, Epoch Jul 2026 Official leaderboard
36 Kimi K2.6 21.7% task version 1.0.0, Epoch Apr 2026 Official leaderboard
37 Grok 4.7 20.8% xhigh; task version 1.0.0, Epoch Sep 2026 Official leaderboard
38 Mistral Large 4 Preview 20.8% high Oct 2026 Epoch AI run
39 Qwen3.8 Max (0902) 20% xhigh; task version 1.0.0, Epoch Sep 2026 Official leaderboard

Epoch AI has run 39 models. Switch the scores above to rank them together.

4 models have held number one on the official leaderboard since its release in September 2026 4 models have held number one across all independent runs since its release in September 2026

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and Epoch AI’s runs.

Anthropic OpenAI
Own leaderboard Passed 85% in Nov 2026, by Claude Fable 5.2
Claude Fable 5.2 90%
Epoch AI runs Passed 85% in Nov 2026, by Claude Fable 5.2
Claude Fable 5.2 90%

Sources: Official Furniture Assembly Benchmark leaderboard and Epoch AI runs (CC BY 4.0)

Claude Fable 5.2 on the official leaderboard passed 85% in November 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Run by Epoch AI Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Random guessing Claude Fable 5.2 Claude Opus 5.5 Claude Fable 5.1 Claude Opus 5 GPT-5.6 Sol GPT-5.5 GPT-5.2 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Random guessing Claude Fable 5.2 Claude Opus 5.5

Sources: Official Furniture Assembly Benchmark leaderboard and Epoch AI runs (CC BY 4.0)

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Vision

All 281 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. GPT-5.1 85.4%
  • MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
  • CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
  • MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
  • MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%
  • ChartQA In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance. Command A Vision 90.9%