Benchmark Almanac

EJBench

GPT-6 Astra 0.71 leads since November 2026
Not in model cards

Models interpret simulation results for battery cooling plates. They match results to test conditions, select evidence for a design decision, and rank operating conditions by temperature.

Gradient Dynamics ranks 12 model configurations on 228 held-out tasks. Each of three task families contributes equally to the score.

Released
November 2026
Kind
Lab and data work
Level
University
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Tables → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Occupations
Architecture and engineering
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
12
Entries on its own leaderboard, 11 models we track

GPT-6 Astra leads with 0.71 GPT-6 Astra leads with 0.71

Best score per entry 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. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

  1. 1 GPT-6 Astra high 0.71
  2. 2 GPT-6.1 Sol high 0.69
  3. 3 Claude Opus 5.5 high 0.64
  4. 4 GPT-6.1 Terra high 0.59
  5. 5 Muse Spark 1.3 high 0.55
  6. 6 Claude Sonnet 5.5 high 0.52
  7. 7 Mistral Large 4 high 0.50
  8. 8 DeepSeek V4.1 Flash high 0.45
  9. 9 Claude Haiku 5.5 high 0.38
  10. 10 GPT-6 Luna high 0.36
  11. 11 Qwen3.5-9B + RL (ours) thinking on, temperature 0.6 0.11
  12. 12 Qwen3.5-9B-Base thinking on, temperature 0.6 0.07

Source: Official EJBench leaderboard

#EntryScore ±SettingReleasedRun onSource
1 GPT-6 Astra 0.71 0.7 to 0.8 high Sep 2026 Nov 2026 Official leaderboard
2 GPT-6.1 Sol 0.69 0.6 to 0.7 high Sep 2026 Nov 2026 Official leaderboard
3 Claude Opus 5.5 0.64 ±0.1 high Sep 2026 Nov 2026 Official leaderboard
4 GPT-6.1 Terra 0.59 0.5 to 0.7 high Oct 2026 Nov 2026 Official leaderboard
5 Muse Spark 1.3 0.55 0.5 to 0.6 high Sep 2026 Nov 2026 Official leaderboard
6 Claude Sonnet 5.5 0.52 0.5 to 0.6 high Sep 2026 Nov 2026 Official leaderboard
7 Mistral Large 4 0.50 0.4 to 0.6 high Oct 2026 Nov 2026 Official leaderboard
8 DeepSeek V4.1 Flash 0.45 0.4 to 0.5 high Sep 2026 Nov 2026 Official leaderboard
9 Claude Haiku 5.5 0.38 ±0.1 high Oct 2026 Nov 2026 Official leaderboard
10 GPT-6 Luna 0.36 0.3 to 0.4 high Sep 2026 Nov 2026 Official leaderboard
11 Qwen3.5-9B + RL (ours) 0.11 0.1 to 0.1 thinking on, temperature 0.6 Nov 2026 Official leaderboard
12 Qwen3.5-9B-Base 0.07 ±0 thinking on, temperature 0.6 Feb 2026 Nov 2026 Official leaderboard
#ModelScore ±SettingReleasedRun onSource
1 GPT-6 Astra 0.71 0.7 to 0.8 high Sep 2026 Nov 2026 Official leaderboard
2 GPT-6.1 Sol 0.69 0.6 to 0.7 high Sep 2026 Nov 2026 Official leaderboard
3 Claude Opus 5.5 0.64 ±0.1 high Sep 2026 Nov 2026 Official leaderboard
4 GPT-6.1 Terra 0.59 0.5 to 0.7 high Oct 2026 Nov 2026 Official leaderboard
5 Muse Spark 1.3 0.55 0.5 to 0.6 high Sep 2026 Nov 2026 Official leaderboard
6 Claude Sonnet 5.5 0.52 0.5 to 0.6 high Sep 2026 Nov 2026 Official leaderboard
7 Mistral Large 4 0.50 0.4 to 0.6 high Oct 2026 Nov 2026 Official leaderboard
8 DeepSeek V4.1 Flash 0.45 0.4 to 0.5 high Sep 2026 Nov 2026 Official leaderboard
9 Claude Haiku 5.5 0.38 ±0.1 high Oct 2026 Nov 2026 Official leaderboard
10 GPT-6 Luna 0.36 0.3 to 0.4 high Sep 2026 Nov 2026 Official leaderboard
11 Qwen3.5-9B-Base 0.07 ±0 thinking on, temperature 0.6 Feb 2026 Nov 2026 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Campaign, Evidence and Envelope. Each ranks models on one part of the test. GPT-6 Astra leads 2 of them, and Muse Spark 1.3 leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Campaign

11 models we track have a score here.

  1. 1 GPT-6 Astra 0.71
  2. 2 GPT-6.1 Sol 0.68
  3. 3 Claude Opus 5.5 0.59
  4. 4 Claude Sonnet 5.5 ↑2 0.54
  5. 5 GPT-6.1 Terra ↓1 0.49
  6. 6 Muse Spark 1.3 ↓1 0.43
  7. 7 DeepSeek V4.1 Flash ↑1 0.35
  8. 8 Mistral Large 4 ↓1 0.35
  9. 9 Claude Haiku 5.5 0.26
  10. 10 GPT-6 Luna 0.18

The best score, 0.71, costs $0.73 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol at high comes within 0 of the best for $0.15 a task, 5 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Anthropic Cost frontier One model at each effort level
0.00 0.20 0.40 0.60 0.80 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-6 Astra, high GPT-6.1 Sol, high GPT-6.1 Terra, high Claude Haiku 5.5, high GPT-6 Luna, high 0.00 0.20 0.40 0.60 0.80 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-6 Astra, high GPT-6.1 Terra, high

Source: Official EJBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra high 0.71 $0.73
GPT-6.1 Sol high 0.69 $0.15
Claude Opus 5.5 high 0.64 $0.43
GPT-6.1 Terra high 0.59 $0.04
Muse Spark 1.3 high 0.55 $0.19
Claude Sonnet 5.5 high 0.52 $0.56
Mistral Large 4 high 0.50 $0.13
DeepSeek V4.1 Flash high 0.45 $0.05
Claude Haiku 5.5 high 0.38 $0.03
GPT-6 Luna high 0.36 $0.007
Qwen3.5-9B-Base thinking on, temperature 0.6 0.07

The best score, 0.71, costs $0.73 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol at high comes within 0 of the best for $0.15 a task, 5 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Anthropic Cost frontier One model at each effort level
0.00 0.20 0.40 0.60 0.80 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-6 Astra, high GPT-6.1 Sol, high GPT-6.1 Terra, high Claude Haiku 5.5, high GPT-6 Luna, high 0.00 0.20 0.40 0.60 0.80 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale GPT-6 Astra, high GPT-6.1 Terra, high

Source: Official EJBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra high 0.71 $0.73
GPT-6.1 Sol high 0.69 $0.15
Claude Opus 5.5 high 0.64 $0.43
GPT-6.1 Terra high 0.59 $0.04
Muse Spark 1.3 high 0.55 $0.19
Claude Sonnet 5.5 high 0.52 $0.56
Mistral Large 4 high 0.50 $0.13
DeepSeek V4.1 Flash high 0.45 $0.05
Claude Haiku 5.5 high 0.38 $0.03
GPT-6 Luna high 0.36 $0.007
Qwen3.5-9B-Base thinking on, temperature 0.6 0.07

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

Source: Official EJBench 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
0.00 0.20 0.40 0.60 0.80 2026 2027 0.00 0.20 0.40 0.60 0.80 2026 2027

Source: Official EJBench 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 Science and reasoning

All 203 in Science and reasoning
  • GPQA Diamond In use 198 graduate-level 'Google-proof' multiple-choice questions in biology, chemistry and physics. Gemini 4 Carbon 97%
  • Terminal-Bench-Science In use 70 scientific research-workflow tasks across five science areas, run in Harbor containers. Hidden tests give the resolution rate. GPT-6 Astra 68.1%
  • BioMysteryBench In use Bioinformatics problems on anonymized real datasets that require real analysis with tools. Rubric-graded, with human-solvable and human-difficult splits. Claude Fable 5.2 83.7%
  • FrontierScience In use Models solve physics, chemistry, and biology problems in an Olympiad track graded by short answer and a Research track graded by 10-point rubrics. Intern-S2-397B 87%
  • HiPhO In use Models solve theory problems from 2024 and 2025 physics olympiad exams, some with diagrams. Official marking schemes grade answers and steps, mapped to medal levels. Gemini 3 Pro 78.3%
  • LABBench2 In use 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. GPT-5.2 60.9%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (12 entries).
  • Pass Models scored 11 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from October 2026, for GPT-6.1 Terra.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 11 rows for models we track and 1 other entry 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