Benchmark Almanac

EEBench

Gemini 4 Carbon 83.9% leads since October 2026
Gemini 4 Carbon 83.9% leads since October 2026
In use

13 electrical-engineering circuit and PCB design tasks graded by deterministic build checks, simulation and BOM constraints. No LLM judge.

Made by
atopile
Released
September 2026
Kind
Electronics
Level
Expert
Scored
Checked automatically
Answers
Code
Strongest model tested
Claude Fable 5.2 #1 on the capability index
Last in a launch post
September 2026
Last model added
November 2026
Scores refreshed
November 12th, 2026
30
Models scored on its own leaderboard
1
Launch post lists it, from 1 lab
3
System cards report it, from 1 lab

Gemini 4 Carbon leads with 83.9%

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

Gemini 4 Carbon leads with 83.9%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

30 models. Scroll sideways for the rest.

  1. 83.9 Gemini 4 Carbon
  2. 77.3 Claude Fable 5.2
  3. 75.0 Claude Opus 5.5
  4. 69.3 GPT-6 Astra
  5. 67.2 Claude Sonnet 5.5
  6. 67.1 Gemini 4 Argon
  7. 67.1 Claude Haiku 5.5
  8. 64.0 Grok 4.7
  9. 63.6 GPT-6.1 Sol
  10. 62.7 Muse Wonder
  11. 61.6 Claude Opus 5
  12. 57.1 Grok 4.6
  13. 56.4 Claude Fable 5.1
  14. 56.3 GPT-6 Sol
  15. 55.4 Gemini 3.8 Flash
  16. 54.3 Claude Fable 5
  17. 51.4 Claude Opus 4.8
  18. 50.9 Grok 4.5
  19. 50.1 Mistral Large 4
  20. 49.4 Muse Spark 1.3
  21. 46.3 Gemini 3.1 Pro
  22. 42.8 Muse Ember
  23. 40.3 Claude Sonnet 5
  24. 37.4 Gemini 3.6 Flash
  25. 34.3 Gemini 3.5 Flash
  26. 31.1 GPT-6 Luna
  27. 20.4 Claude Sonnet 4.6
  28. 19.5 Gemini 3 Flash
  29. 18.0 Grok Build 0.1
  30. 3.50 Claude Haiku 4.5
  1. 1 Gemini 4 Carbon high 83.9%
  2. 2 Claude Fable 5.2 xhigh 77.3%
  3. 3 Claude Opus 5.5 xhigh 75%
  4. 4 GPT-6 Astra 69.3%
  5. 5 Claude Sonnet 5.5 xhigh 67.2%
  6. 6 Gemini 4 Argon high 67.1%
  7. 7 Claude Haiku 5.5 high 67.1%
  8. 8 Grok 4.7 64%
  9. 9 GPT-6.1 Sol high 63.6%
  10. 10 Muse Wonder high 62.7%
  11. 11 Claude Opus 5 61.6%
  12. 12 Grok 4.6 57.1%
  13. 13 Claude Fable 5.1 high 56.4%
  14. 14 GPT-6 Sol xhigh 56.3%
  15. 15 Gemini 3.8 Flash high 55.4%
  16. 16 Claude Fable 5 54.3%
  17. 17 Claude Opus 4.8 max 51.4%
  18. 18 Grok 4.5 50.9%
  19. 19 Mistral Large 4 high 50.1%
  20. 20 Muse Spark 1.3 high 49.4%
  21. 21 Gemini 3.1 Pro 46.3%
  22. 22 Muse Ember max 42.8%
  23. 23 Claude Sonnet 5 40.3%
  24. 24 Gemini 3.6 Flash 37.4%
  25. 25 Gemini 3.5 Flash 34.3%
  26. 26 GPT-6 Luna high 31.1%
  27. 27 Claude Sonnet 4.6 20.4%
  28. 28 Gemini 3 Flash 19.5%
  29. 29 Grok Build 0.1 18%
  30. 30 Claude Haiku 4.5 3.50%

Source: Official EEBench leaderboard

30 models. Scroll sideways for the rest.

  1. 83.9 Gemini 4 Carbon
  2. 77.3 Claude Fable 5.2
  3. 75.0 Claude Opus 5.5
  4. 69.3 GPT-6 Astra
  5. 67.2 Claude Sonnet 5.5
  6. 67.1 Gemini 4 Argon
  7. 67.1 Claude Haiku 5.5
  8. 66.0 Grok 4.7
  9. 63.6 GPT-6.1 Sol
  10. 62.7 Muse Wonder
  11. 61.6 Claude Opus 5
  12. 60.0 Grok 4.6
  13. 56.4 Claude Fable 5.1
  14. 56.3 GPT-6 Sol
  15. 55.4 Gemini 3.8 Flash
  16. 54.3 Claude Fable 5
  17. 51.4 Claude Opus 4.8
  18. 50.9 Grok 4.5
  19. 50.1 Mistral Large 4
  20. 49.4 Muse Spark 1.3
  21. 46.3 Gemini 3.1 Pro
  22. 42.8 Muse Ember
  23. 40.3 Claude Sonnet 5
  24. 37.4 Gemini 3.6 Flash
  25. 34.3 Gemini 3.5 Flash
  26. 31.1 GPT-6 Luna
  27. 20.4 Claude Sonnet 4.6
  28. 19.5 Gemini 3 Flash
  29. 18.0 Grok Build 0.1
  30. 3.50 Claude Haiku 4.5
  1. 1 Gemini 4 Carbon high 83.9%
  2. 2 Claude Fable 5.2 xhigh 77.3%
  3. 3 Claude Opus 5.5 xhigh 75%
  4. 4 GPT-6 Astra 69.3%
  5. 5 Claude Sonnet 5.5 xhigh 67.2%
  6. 6 Gemini 4 Argon high 67.1%
  7. 7 Claude Haiku 5.5 high 67.1%
  8. 8 Grok 4.7 66%
  9. 9 GPT-6.1 Sol high 63.6%
  10. 10 Muse Wonder high 62.7%
  11. 11 Claude Opus 5 61.6%
  12. 12 Grok 4.6 60%
  13. 13 Claude Fable 5.1 high 56.4%
  14. 14 GPT-6 Sol xhigh 56.3%
  15. 15 Gemini 3.8 Flash high 55.4%
  16. 16 Claude Fable 5 54.3%
  17. 17 Claude Opus 4.8 max 51.4%
  18. 18 Grok 4.5 50.9%
  19. 19 Mistral Large 4 high 50.1%
  20. 20 Muse Spark 1.3 high 49.4%
  21. 21 Gemini 3.1 Pro 46.3%
  22. 22 Muse Ember max 42.8%
  23. 23 Claude Sonnet 5 40.3%
  24. 24 Gemini 3.6 Flash 37.4%
  25. 25 Gemini 3.5 Flash 34.3%
  26. 26 GPT-6 Luna high 31.1%
  27. 27 Claude Sonnet 4.6 20.4%
  28. 28 Gemini 3 Flash 19.5%
  29. 29 Grok Build 0.1 18%
  30. 30 Claude Haiku 4.5 3.50%

Sources: Official EEBench leaderboard and labs’ model cards

#ModelScoreSettingDateSource
1 Gemini 4 Carbon 83.9% high, Gemini CLI Oct 2026 Official leaderboard
2 Claude Fable 5.2 77.3% xhigh, Claude Code Nov 2026 Official leaderboard
3 Claude Opus 5.5 75% xhigh, Claude Code Sep 2026 Official leaderboard
4 GPT-6 Astra 69.3% Codex CLI Sep 2026 Official leaderboard
5 Claude Sonnet 5.5 67.2% xhigh, Claude Code Sep 2026 Official leaderboard
6 Gemini 4 Argon 67.1% high, Gemini CLI Sep 2026 Official leaderboard
7 Claude Haiku 5.5 67.1% high, Claude Code Oct 2026 Official leaderboard
8 Grok 4.7 64% Grok CLI Sep 2026 Official leaderboard
9 GPT-6.1 Sol 63.6% high, Codex CLI Sep 2026 Official leaderboard
10 Muse Wonder 62.7% high, Muse Code Oct 2026 Official leaderboard
11 Claude Opus 5 61.6% Claude Code Jul 2026 Official leaderboard
12 Grok 4.6 57.1% Grok CLI Aug 2026 Official leaderboard
13 Claude Fable 5.1 56.4% high, Claude Code Sep 2026 Official leaderboard
14 GPT-6 Sol 56.3% xhigh, Codex CLI Sep 2026 Official leaderboard
15 Gemini 3.8 Flash 55.4% high, Gemini CLI Sep 2026 Official leaderboard
16 Claude Fable 5 54.3% Claude Code Jun 2026 Official leaderboard
17 Claude Opus 4.8 51.4% max, Claude Code May 2026 Official leaderboard
18 Grok 4.5 50.9% Grok CLI Jul 2026 Official leaderboard
19 Mistral Large 4 50.1% high, Mistral Vibe Oct 2026 Official leaderboard
20 Muse Spark 1.3 49.4% high, Muse Code Sep 2026 Official leaderboard
21 Gemini 3.1 Pro 46.3% Gemini CLI Feb 2026 Official leaderboard
22 Muse Ember 42.8% max, Muse Code Nov 2026 Official leaderboard
23 Claude Sonnet 5 40.3% Claude Code Jun 2026 Official leaderboard
24 Gemini 3.6 Flash 37.4% Gemini CLI Jul 2026 Official leaderboard
25 Gemini 3.5 Flash 34.3% Gemini CLI May 2026 Official leaderboard
26 GPT-6 Luna 31.1% high, Codex CLI Sep 2026 Official leaderboard
27 Claude Sonnet 4.6 20.4% Claude Code Feb 2026 Official leaderboard
28 Gemini 3 Flash 19.5% Gemini CLI Dec 2025 Official leaderboard
29 Grok Build 0.1 18% Grok CLI May 2026 Official leaderboard
30 Claude Haiku 4.5 3.50% Claude Code Oct 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 4 Carbon 83.9% high, Gemini CLI Oct 2026 Official leaderboard
2 Claude Fable 5.2 77.3% xhigh, Claude Code Nov 2026 Official leaderboard
3 Claude Opus 5.5 75% xhigh, Claude Code Sep 2026 Official leaderboard
4 GPT-6 Astra 69.3% Codex CLI Sep 2026 Official leaderboard
5 Claude Sonnet 5.5 67.2% xhigh, Claude Code Sep 2026 Official leaderboard
6 Gemini 4 Argon 67.1% high, Gemini CLI Sep 2026 Official leaderboard
7 Claude Haiku 5.5 67.1% high, Claude Code Oct 2026 Official leaderboard
8 Grok 4.7 66% Sep 2026 Lab model card
9 GPT-6.1 Sol 63.6% high, Codex CLI Sep 2026 Official leaderboard
10 Muse Wonder 62.7% high, Muse Code Oct 2026 Official leaderboard
11 Claude Opus 5 61.6% Claude Code Jul 2026 Official leaderboard
12 Grok 4.6 60% Aug 2026 Lab model card
13 Claude Fable 5.1 56.4% high, Claude Code Sep 2026 Official leaderboard
14 GPT-6 Sol 56.3% xhigh, Codex CLI Sep 2026 Official leaderboard
15 Gemini 3.8 Flash 55.4% high, Gemini CLI Sep 2026 Official leaderboard
16 Claude Fable 5 54.3% Claude Code Jun 2026 Official leaderboard
17 Claude Opus 4.8 51.4% max, Claude Code May 2026 Official leaderboard
18 Grok 4.5 50.9% Grok CLI Jul 2026 Official leaderboard
19 Mistral Large 4 50.1% high, Mistral Vibe Oct 2026 Official leaderboard
20 Muse Spark 1.3 49.4% high, Muse Code Sep 2026 Official leaderboard
21 Gemini 3.1 Pro 46.3% Gemini CLI Feb 2026 Official leaderboard
22 Muse Ember 42.8% max, Muse Code Nov 2026 Official leaderboard
23 Claude Sonnet 5 40.3% Claude Code Jun 2026 Official leaderboard
24 Gemini 3.6 Flash 37.4% Gemini CLI Jul 2026 Official leaderboard
25 Gemini 3.5 Flash 34.3% Gemini CLI May 2026 Official leaderboard
26 GPT-6 Luna 31.1% high, Codex CLI Sep 2026 Official leaderboard
27 Claude Sonnet 4.6 20.4% Claude Code Feb 2026 Official leaderboard
28 Gemini 3 Flash 19.5% Gemini CLI Dec 2025 Official leaderboard
29 Grok Build 0.1 18% Grok CLI May 2026 Official leaderboard
30 Claude Haiku 4.5 3.50% Claude Code Oct 2025 Official leaderboard

Labs report scores for 2 models in their cards. Switch the scores above to rank them together.

Benchmark score versus per-task metrics

The board switches the x-axis between cost, tokens, steps and time. This is its Time view: mean seconds per attempt. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0% 20% 40% 60% 80% 100% 01,0002,0003,000 Time per attempt (s), lower is better Score Claude Haiku 4.5: Score 3.5%, time per attempt 344 s Gemini 3 Flash: Score 19.5%, time per attempt 437 s Claude Sonnet 4.6: Score 20.4%, time per attempt 1,439 s Gemini 3.1 Pro: Score 46.3%, time per attempt 339 s Gemini 3.5 Flash: Score 34.3%, time per attempt 452 s Claude Opus 4.8 (max): Score 51.4%, time per attempt 2,159 s Claude Opus 4.8 (high): Score 35.2%, time per attempt 512 s Grok Build 0.1: Score 18%, time per attempt 669 s Claude Fable 5: Score 54.3%, time per attempt 564 s Claude Sonnet 5: Score 40.3%, time per attempt 1,198 s Grok 4.5: Score 50.9%, time per attempt 372 s Gemini 3.6 Flash: Score 37.4%, time per attempt 797 s Claude Opus 5: Score 61.6%, time per attempt 781 s Grok 4.6: Score 57.1%, time per attempt 2,126 s Claude Fable 5.1 (high): Score 56.4%, time per attempt 652 s Gemini 3.8 Flash (high): Score 55.4%, time per attempt 587 s Muse Spark 1.3 (high): Score 49.4%, time per attempt 423 s Grok 4.7: Score 64%, time per attempt 1,234 s Claude Opus 5.5 (high): Score 68.9%, time per attempt 247 s GPT-6 Luna (high): Score 31.1%, time per attempt 388 s GPT-6 Sol (xhigh): Score 56.3%, time per attempt 359 s Claude Sonnet 5.5 (xhigh): Score 67.2%, time per attempt 455 s GPT-6.1 Sol (high): Score 63.6%, time per attempt 317 s Gemini 4 Argon (high): Score 67.1%, time per attempt 1,136 s Mistral Large 4 (high): Score 50.1%, time per attempt 2,019 s Claude Haiku 5.5 (high): Score 67.1%, time per attempt 917 s Muse Wonder (high): Score 62.7%, time per attempt 245 s Muse Ember (max): Score 42.8%, time per attempt 317 s GPT-6 Astra: Score 69.3%, time per attempt 106 s Claude Opus 5.5 (xhigh): Score 75%, time per attempt 802 s Gemini 4 Carbon (high): Score 83.9%, time per attempt 2,987 s Claude Fable 5.2 (xhigh): Score 77.3%, time per attempt 1,772 s Gemini 4 Carbon (high)Claude Fable 5.2 (xhigh)Claude Opus 5.5 (xhigh)GPT-6 Astra 0% 20% 40% 60% 80% 100% 01,0002,0003,000 Time per attempt (s), lower is better Score Claude Haiku 4.5: Score 3.5%, time per attempt 344 s Gemini 3 Flash: Score 19.5%, time per attempt 437 s Claude Sonnet 4.6: Score 20.4%, time per attempt 1,439 s Gemini 3.1 Pro: Score 46.3%, time per attempt 339 s Gemini 3.5 Flash: Score 34.3%, time per attempt 452 s Claude Opus 4.8 (max): Score 51.4%, time per attempt 2,159 s Claude Opus 4.8 (high): Score 35.2%, time per attempt 512 s Grok Build 0.1: Score 18%, time per attempt 669 s Claude Fable 5: Score 54.3%, time per attempt 564 s Claude Sonnet 5: Score 40.3%, time per attempt 1,198 s Grok 4.5: Score 50.9%, time per attempt 372 s Gemini 3.6 Flash: Score 37.4%, time per attempt 797 s Claude Opus 5: Score 61.6%, time per attempt 781 s Grok 4.6: Score 57.1%, time per attempt 2,126 s Claude Fable 5.1 (high): Score 56.4%, time per attempt 652 s Gemini 3.8 Flash (high): Score 55.4%, time per attempt 587 s Muse Spark 1.3 (high): Score 49.4%, time per attempt 423 s Grok 4.7: Score 64%, time per attempt 1,234 s Claude Opus 5.5 (high): Score 68.9%, time per attempt 247 s GPT-6 Luna (high): Score 31.1%, time per attempt 388 s GPT-6 Sol (xhigh): Score 56.3%, time per attempt 359 s Claude Sonnet 5.5 (xhigh): Score 67.2%, time per attempt 455 s GPT-6.1 Sol (high): Score 63.6%, time per attempt 317 s Gemini 4 Argon (high): Score 67.1%, time per attempt 1,136 s Mistral Large 4 (high): Score 50.1%, time per attempt 2,019 s Claude Haiku 5.5 (high): Score 67.1%, time per attempt 917 s Muse Wonder (high): Score 62.7%, time per attempt 245 s Muse Ember (max): Score 42.8%, time per attempt 317 s GPT-6 Astra: Score 69.3%, time per attempt 106 s Claude Opus 5.5 (xhigh): Score 75%, time per attempt 802 s Gemini 4 Carbon (high): Score 83.9%, time per attempt 2,987 s Claude Fable 5.2 (xhigh): Score 77.3%, time per attempt 1,772 s Gemini 4 Carbon (high)Claude Fable 5.2 (xhigh)Claude Opus 5.5 (xhigh)GPT-6 Astra
ModelSettingScoreTime per attempt
Gemini 4 Carbon high, Gemini CLI 83.9% 2,987 s
Claude Fable 5.2 xhigh, Claude Code 77.3% 1,772 s
Claude Opus 5.5 xhigh, Claude Code 75% 802 s
GPT-6 Astra Codex CLI 69.3% 106 s
Claude Opus 5.5 high, Claude Code 68.9% 247 s
Claude Sonnet 5.5 xhigh, Claude Code 67.2% 455 s
Claude Haiku 5.5 high, Claude Code 67.1% 917 s
Gemini 4 Argon high, Gemini CLI 67.1% 1,136 s
Grok 4.7 Grok CLI 64% 1,234 s
GPT-6.1 Sol high, Codex CLI 63.6% 317 s
Muse Wonder high, Muse Code 62.7% 245 s
Claude Opus 5 Claude Code 61.6% 781 s
Grok 4.6 Grok CLI 57.1% 2,126 s
Claude Fable 5.1 high, Claude Code 56.4% 652 s
GPT-6 Sol xhigh, Codex CLI 56.3% 359 s
Gemini 3.8 Flash high, Gemini CLI 55.4% 587 s
Claude Fable 5 Claude Code 54.3% 564 s
Claude Opus 4.8 max, Claude Code 51.4% 2,159 s
Grok 4.5 Grok CLI 50.9% 372 s
Mistral Large 4 high, Mistral Vibe 50.1% 2,019 s
Muse Spark 1.3 high, Muse Code 49.4% 423 s
Gemini 3.1 Pro Gemini CLI 46.3% 339 s
Muse Ember max, Muse Code 42.8% 317 s
Claude Sonnet 5 Claude Code 40.3% 1,198 s
Gemini 3.6 Flash Gemini CLI 37.4% 797 s
Claude Opus 4.8 high, Claude Code 35.2% 512 s
Gemini 3.5 Flash Gemini CLI 34.3% 452 s
GPT-6 Luna high, Codex CLI 31.1% 388 s
Claude Sonnet 4.6 Claude Code 20.4% 1,439 s
Gemini 3 Flash Gemini CLI 19.5% 437 s
Grok Build 0.1 Grok CLI 18% 669 s
Claude Haiku 4.5 Claude Code 3.5% 344 s

The best score, 83.9%, costs $6.11 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Anthropic Google DeepMind Meta Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Gemini 4 Carbon, high Claude Opus 5.5, xhigh GPT-6 Astra GPT-6.1 Sol, high Muse Ember, max GPT-6 Luna, high 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale Gemini 4 Carbon, high Claude Opus 5.5, xhigh

Source: Official EEBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 4 Carbon high Gemini CLI 83.9% $6.11 124,587
Claude Fable 5.2 xhigh Claude Code 77.3% $8.87 139,765
Claude Opus 5.5 xhigh Claude Code 75% $2.37 65,077
GPT-6 Astra Codex CLI 69.3% $0.83 5,967
Claude Opus 5.5 high Claude Code 68.9% $0.65 20,320
Claude Sonnet 5.5 xhigh Claude Code 67.2% $0.83 44,780
Gemini 4 Argon high Gemini CLI 67.1% $1.90 73,041
Claude Haiku 5.5 high Claude Code 67.1% $0.71 150,156
Grok 4.7 Grok CLI 64% $3.55 81,243
GPT-6.1 Sol high Codex CLI 63.6% $0.30 17,181
Muse Wonder high Muse Code 62.7% $1.14 27,725
Claude Opus 5 Claude Code 61.6% $2.38 49,030
Grok 4.6 Grok CLI 57.1% $2.11 126,635
Claude Fable 5.1 high Claude Code 56.4% $2.92 40,465
GPT-6 Sol xhigh Codex CLI 56.3% $0.65 26,759
Gemini 3.8 Flash high Gemini CLI 55.4% $1.38 68,165
Claude Fable 5 Claude Code 54.3% $3.19 33,224
Claude Opus 4.8 max Claude Code 51.4% $5.73 126,821
Grok 4.5 Grok CLI 50.9% $1.14 27,806
Mistral Large 4 high Mistral Vibe 50.1% $1.43 129,780
Muse Spark 1.3 high Muse Code 49.4% $0.84 48,389
Gemini 3.1 Pro Gemini CLI 46.3% $0.84 28,279
Muse Ember max Muse Code 42.8% $0.12 86,712
Claude Sonnet 5 Claude Code 40.3% $4.21 85,208
Gemini 3.6 Flash Gemini CLI 37.4% $1.92 47,303
Claude Opus 4.8 high Claude Code 35.2% $1.44 34,450
Gemini 3.5 Flash Gemini CLI 34.3% $1.11 41,102
GPT-6 Luna high Codex CLI 31.1% $0.04 33,945
Claude Sonnet 4.6 Claude Code 20.4% $1.89 80,080
Gemini 3 Flash Gemini CLI 19.5% $0.40 53,382
Grok Build 0.1 Grok CLI 18% $1.66 56,810
Claude Haiku 4.5 Claude Code 3.50% $0.41 31,601

4 models have held number one on the official leaderboard since its release in September 2026 4 models have held number one with claims included 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 what labs claim in their cards.

Anthropic OpenAI Google DeepMind xAI
Own leaderboard 4 leaders
Gemini 4 Carbon 83.9%
Lab claims Not tested since Sep 2026
Grok 4.7 66%

Sources: Official EEBench leaderboard and labs’ model cards

Scores over time

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 Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Carbon Claude Opus 5.5 GPT-6 Astra Claude Opus 5 Claude Fable 5 Gemini 3.1 Pro Claude Sonnet 4.6 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Carbon Claude Opus 5.5

Sources: Official EEBench leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

xAI
DateLaunch postScoreAs the post writes it
Sep 21, 2026 Grok 4.7 64 EEBench (Electrical engineering)

3 system cards report it, from 1 lab

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Sep 21, 2026 Grok 4.7 model card 66 EEBench V1 core corpus, reward, xhigh, Grok Build harness, run by Atopile (4.1) Capabilities
Aug 12, 2026 Grok 4.6 model card 60 EEBench V1 core corpus, reward, xhigh, Grok Build harness, run by Atopile (4.1). High = 53.0% Capabilities
Jul 14, 2026 Grok 4.5 model card EEBench V1 core corpus, score vs output tokens scatter, no printed value for Grok 4.5 (3.1) Capabilities

The lab's number comes out ahead in 2 of 2 pairs that differ

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.

Model Lab says Board says Gap
Grok 4.6 60% EEBench V1 core corpus, reward, xhigh, Grok Build harness, run by Atopile (4.1). High = 53.0% 57.1% Leaderboard, Grok CLI +2.9 pts
Grok 4.7 66% EEBench V1 core corpus, reward, xhigh, Grok Build harness, run by Atopile (4.1) 64% Leaderboard, Grok CLI +2.0 pts

More in Engineering and CAD

All 23 in Engineering and CAD
  • Parametric CAD Bench In use Agents author editable parametric FreeCAD part designs, graded against held-back references for geometry and spec consistency. Claude Opus 5.5 61%
  • 3DCodeBench In use Agentic procedural 3D modeling via code: author engine-ready 3D assets through software APIs. Graded for executability and shape fidelity. LongCat-2.5 55.4%
  • CADGenBench Only in system cards Agents construct CAD models from design prompts. The generation split is graded for executability and geometric correctness. Claude Fable 5 45.1%
  • PartBench Only in system cards Parametric mechanical part design and CAD-style generation, graded for geometric and manufacturing fidelity. Grok 4.6 58.5%
  • LMBuild Not in model cards Language model agents design 3D structures as parts, joints, materials and assembly steps. Scoring checks that the structures can be built and work as intended. GPT-6 Astra 66.5%
  • Physical-Engineering's Last Exam Not in model cards Agents produce billable engineering deliverables such as calculation packages, drawings and design reviews in 85 tasks across electrical, civil, structural and mechanical engineering. Automatic graders check constraints, safety limits and the design objective. Claude Opus 5.5 0.49