Benchmark Almanac

AutoFactory-Bench

GPT-6 Astra 72.2% leads since October 2026
Not in model cards

Agents write control programs for simulated factory cells. A run passes when the program completes every operating condition without a safety violation.

The leaderboard ranks agents on 30 hard tasks. The team publishes 10 of the 100 tasks and keeps the verifier private.

Released
October 2026
Kind
Electronics
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Measures
Capability
Language
English
In and out
Text → Code
Interaction
Agentic
Environment
A simulator
Refresh
One fixed set
Run by
A host
In views
Agents
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
11
Models scored on its own leaderboard

GPT-6 Astra leads with 72.2%

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

  1. 1 GPT-6 Astra 72.2%
  2. 2 GPT-6.1 Sol 70%
  3. 3 Claude Opus 5.5 60.3%
  4. 4 GPT-6.1 Terra 57.8%
  5. 5 Claude Fable 5.1 55.6%
  6. 6 Claude Sonnet 5.5 42.7%
  7. 7 DeepSeek V4.1 Flash OpenGateway 21.1%
  8. 8 Claude Haiku 5.5 11.1%
  9. 9 GPT-6 Luna 8.89%
  10. 10 Nemotron 3.5 Super VL OpenRouter (DeepInfra) 5.56%
  11. 11 GLM-5.3-Flash OpenGateway 4.44%

Source: Official AutoFactory-Bench leaderboard

#ModelScore ±SettingDateSource
1 GPT-6 Astra 72.2% 57.8 to 85.6 Codex Sep 2026 Official leaderboard
2 GPT-6.1 Sol 70% 55.6 to 83.3 Codex Sep 2026 Official leaderboard
3 Claude Opus 5.5 60.3% 48.9 to 71.1 Claude Code Sep 2026 Official leaderboard
4 GPT-6.1 Terra 57.8% 42.2 to 74.4 Codex Oct 2026 Official leaderboard
5 Claude Fable 5.1 55.6% ±14.4 Claude Code Sep 2026 Official leaderboard
6 Claude Sonnet 5.5 42.7% 28.9 to 57.1 Claude Code Sep 2026 Official leaderboard
7 DeepSeek V4.1 Flash 21.1% 10 to 34.4 OpenGateway, mini-swe-agent Sep 2026 Official leaderboard
8 Claude Haiku 5.5 11.1% 3.3 to 21.1 Claude Code Oct 2026 Official leaderboard
9 GPT-6 Luna 8.89% 2.2 to 17.8 Codex Sep 2026 Official leaderboard
10 Nemotron 3.5 Super VL 5.56% 0 to 12.2 OpenRouter (DeepInfra), mini-swe-agent Oct 2026 Official leaderboard
11 GLM-5.3-Flash 4.44% 1.1 to 8.9 OpenGateway, mini-swe-agent Aug 2026 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Easy, Medium and All 100 tasks. Each ranks models on one part of the test. GPT-6 Astra leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.

Easy

11 models we track have a score here.

  1. 1 GPT-6 Astra 100%
  2. 2 GPT-6.1 Sol 100%
  3. 3 GPT-6.1 Terra ↑1 96.7%
  4. 4 Claude Fable 5.1 ↑1 92.2%
  5. 5 DeepSeek V4.1 Flash ↑2 92.2%
  6. 6 Claude Opus 5.5 ↓3 92.2%
  7. 7 Claude Haiku 5.5 ↑1 90%
  8. 8 Nemotron 3.5 Super VL ↑2 86.7%
  9. 9 Claude Sonnet 5.5 ↓3 85.6%
  10. 10 GLM-5.3-Flash ↑1 77.8%

The best score, 72.2%, costs $0.98 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol comes within 2.2 points of the best for $0.25 a task, 4 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI DeepSeek 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 GPT-6 Astra GPT-6.1 Sol GPT-6.1 Terra DeepSeek V4.1 Flash, OpenGateway GPT-6 Luna 0% 20% 40% 60% 80% 100% $0.01 $0.10 $1.00 $10 Cost per task, log scale GPT-6 Astra GPT-6.1 Terra

Source: Official AutoFactory-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra Codex 72.2% $0.98
GPT-6.1 Sol Codex 70% $0.25
Claude Opus 5.5 Claude Code 60.3% $0.87
GPT-6.1 Terra Codex 57.8% $0.04
Claude Fable 5.1 Claude Code 55.6% $5.61
Claude Sonnet 5.5 Claude Code 42.7% $0.39
DeepSeek V4.1 Flash OpenGateway mini-swe-agent 21.1% $0.02
Claude Haiku 5.5 Claude Code 11.1% $0.05
GPT-6 Luna Codex 8.89% $0.01
Nemotron 3.5 Super VL OpenRouter (DeepInfra) mini-swe-agent 5.56% $0.07
GLM-5.3-Flash OpenGateway mini-swe-agent 4.44% $0.02

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard 1 leader
GPT-6 Astra 72.2%

Source: Official AutoFactory-Bench leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) GPT-6 Astra Claude Fable 5.1 GLM-5.3-Flash 0% 20% 40% 60% 80% 100% 2027 Near the ceiling (85%) GPT-6 Astra Claude Fable 5.1

Source: Official AutoFactory-Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Engineering and CAD

All 112 in Engineering and CAD
  • EEBench In use 13 electrical-engineering circuit and PCB design tasks graded by deterministic build checks, simulation and BOM constraints. No LLM judge. Gemini 4 Carbon 83.9%
  • 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%
  • Razavi-Bench Not in model cards Models explain what happens in analog circuits and why. A language model awards up to four points per answer against a reference and a rubric. Gemini 4 Carbon 98.3%
  • CircuitRubric Not in model cards Models write analog circuit netlists from design requests. A program compares transistor connections and relative device sizes with accepted reference circuits. Claude Opus 4.8 96%

Data audit

All 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (11 entries).
  • Pass Models scored 11 models we track.
  • Pass Newest score The newest is from October 2026, for GPT-6.1 Terra.
  • Pass Frontier models 6 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 72.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on autofactorybench.com: 11 rows for models we track at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check