Benchmark Almanac

SCI-AGENT-EXAM

GPT-6 Astra 80.9 leads since October 2026
Not in model cards

Agents complete research tasks across 11 scientific fields. Programs check executable outputs, and a model judge grades written work.

Released
October 2026
Kind
Research science
Level
Expert
Scored
Graded by a model
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Code, Text, Actions
Interaction
Agentic
Environment
Terminal, Code sandbox
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
10
Models scored on its own leaderboard

GPT-6 Astra leads with 80.9 among models released by September 2026

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.

The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-6 Astra 80.9
  2. 2 Claude Fable 5 67.3
  3. 3 Claude Opus 5 64.2
  4. 4 Qwen 3.8 Max 62.7
  5. 5 Kimi K3 62.6
  6. 6 GPT-5.6 Sol 59.6
  7. 7 Claude Opus 4.8 57.1
  8. 8 GLM-5.3 55.9
  9. 9 Grok 4.6 52.4
  10. 10 DeepSeek-V4-Pro 50.9

Source: Official SCI-AGENT-EXAM leaderboard

#ModelScoreSettingDateSource
1 GPT-6 Astra 80.9 Sep 2026 Official leaderboard
2 Claude Fable 5 67.3 Jun 2026 Official leaderboard
3 Claude Opus 5 64.2 Jul 2026 Official leaderboard
4 Qwen 3.8 Max 62.7 Aug 2026 Official leaderboard
5 Kimi K3 62.6 Jul 2026 Official leaderboard
6 GPT-5.6 Sol 59.6 Jul 2026 Official leaderboard
7 Claude Opus 4.8 57.1 May 2026 Official leaderboard
8 GLM-5.3 55.9 Aug 2026 Official leaderboard
9 Grok 4.6 52.4 Aug 2026 Official leaderboard
10 DeepSeek-V4-Pro 50.9 Apr 2026 Official leaderboard

Rankings by part

We read 18 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Physical Science, Physics Simulation and Scientific Computation. GPT-6 Astra leads 17 of them, and Claude Fable 5 leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Physical Science

10 models we track have a score here.

  1. 1 GPT-6 Astra 82.8
  2. 2 Claude Opus 5 ↑1 68.9
  3. 3 Claude Fable 5 ↓1 55.4
  4. 4 Claude Opus 4.8 ↑3 53.2
  5. 5 GPT-5.6 Sol ↑1 52.8
  6. 6 Qwen 3.8 Max ↓2 50.1
  7. 7 Kimi K3 ↓2 49.9
  8. 8 GLM-5.3 47.5
  9. 9 DeepSeek-V4-Pro ↑1 40.7
  10. 10 Grok 4.6 ↓1 36.0

The best score, 80.9, costs $21 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.

Anthropic OpenAI Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 100.0 $1.00 $2.00 $5.00 $10 $20 $50 $100 Cost per task, log scale GPT-6 Astra Claude Fable 5 Claude Opus 4.8 0.00 20.0 40.0 60.0 80.0 100.0 $1.00 $10 $100 Cost per task, log scale GPT-6 Astra Claude Fable 5

Source: Official SCI-AGENT-EXAM leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra 80.9 $21 28,583
Claude Fable 5 67.3 $7.67 51,087
Claude Opus 5 64.2 $12 67,797
Qwen 3.8 Max 62.7 $22 114,405
Kimi K3 62.6 $21 63,987
GPT-5.6 Sol 59.6 $13 21,638
Claude Opus 4.8 57.1 $5.13 57,566
GLM-5.3 55.9 $58 98,254
Grok 4.6 52.4 $13 100,347
DeepSeek-V4-Pro 50.9 $32 95,669

Number one over time

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

OpenAI
Own leaderboard Not tested since Oct 2026

Source: Official SCI-AGENT-EXAM 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
50.0 60.0 70.0 80.0 90.0 2026 50.0 60.0 70.0 80.0 90.0 2026

Source: Official SCI-AGENT-EXAM 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 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (10 entries).
  • Pass Models scored 10 models we track.
  • Pass Newest score The newest is from September 2026, for GPT-6 Astra.
  • Note Frontier models Only 1 of the top 10 has a score. The strongest tested is 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 skylenage-ai.github.io: 10 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