Benchmark Almanac

HEPToolBench 1.2

Gemini 3.1 Flash-Lite 71.4% leads since August 2026
Not in model cards

Models generate inputs for particle physics simulations and inspect the resulting files. The main suite has 28 tasks and a separate extension adds three debugging tasks.

The corrected leaderboard lists 42 deployments. Two deployments use Llama 3.3 70B. Version 1.2.1 corrects scores on the same tasks.

Released
August 2026
Kind
Lab and data work
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Code, Actions, Text
Interaction
Agentic
Environment
Terminal, A simulator
Run by
Anyone
In views
Agents, Robotics
Strongest model tested
Gemini 3.5 Flash #44 on the capability index
Last model added
June 2026
Scores refreshed
November 20th, 2026
42
Entries on its own leaderboard, 29 models we track

Gemini 3.1 Flash-Lite leads with 71.4% among models released by June 2026 Gemini 3.1 Flash-Lite leads with 71.4% among models released by June 2026

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. 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.

The newest model with a score here came out in June 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 Gemini 3.1 Flash-Lite Google API / api_cloud 71.4%
  2. 2 GPT-4.1 GitHub Models API / api_cloud 67.9%
  3. 3 gpt-oss-120b Ollama/local / local_ollama 67.9%
  4. 4 GPT-4.1 mini GitHub Models API / api_cloud 64.3%
  5. 5 Llama 3.3 70B (API) GitHub Models API / api_cloud 57.1%
  6. 6 Qwen3-Next-80B-A3B Ollama/local / local_ollama 57.1%
  7. 7 Gemini 3.5 Flash Google API / api_cloud 57.1%
  8. 8 Magistral Medium 2509 Mistral API / api_cloud 53.6%
  9. 9 Llama 3.3 70B (local) Ollama/local / local_ollama 50%
  10. 10 Qwen3-14B Ollama/local / local_ollama 50%
  11. 11 gpt-oss-20b Ollama/local / local_ollama 50%
  12. 12 Gemma 4 26B A4B Ollama/local / local_ollama 50%
  13. 13 Gemma 4 31B IT Ollama/local / local_ollama 50%
  14. 14 DeepSeek-R1 70B Ollama/local / local_ollama 46.4%
  15. 15 Mistral Large 2512 Mistral API / api_cloud 46.4%
  16. 16 Mistral Medium 2508 Mistral API / api_cloud 46.4%
  17. 17 GPT-4.1 nano GitHub Models API / api_cloud 46.4%
  18. 18 Qwen3-Coder-Next Ollama/local / local_ollama 46.4%
  19. 19 Qwen3.5-27B Ollama/local / local_ollama 46.4%
  20. 20 Devstral 2 Mistral API / api_cloud 42.9%
  21. 21 Sarvam 105B Sarvam API / api_cloud 42.9%
  22. 22 Gemma 4 12B Ollama/local / local_ollama 42.9%
  23. 23 Codestral 2508 Mistral API / api_cloud 39.3%
  24. 24 Falcon 3 10B Ollama/local / local_ollama 39.3%
  25. 25 Phi-4 14B Ollama/local / local_ollama 39.3%
  26. 26 Gemma 3 12B Ollama/local / local_ollama 39.3%
  27. 27 Devstral 24B Ollama/local / local_ollama 35.7%
  28. 28 Qwen3-8B Ollama/local / local_ollama 35.7%
  29. 29 Qwen2.5-Coder-14B Ollama/local / local_ollama 32.1%
  30. 30 Ministral 3 14B Ollama/local / local_ollama 32.1%
  31. 31 Gemma 4 E4B Ollama/local / local_ollama 32.1%
  32. 32 Granite 4 32B-A9B Ollama/local / local_ollama 28.6%
  33. 33 Mistral Small 3.2 24B Ollama/local / local_ollama 28.6%
  34. 34 Qwen3.5-35B-A3B Ollama/local / local_ollama 28.6%
  35. 35 Llama 3-8B Ollama/local / local_ollama 21.4%
  36. 36 Llama 3.1-8B Ollama/local / local_ollama 21.4%
  37. 37 Qwen2.5-Coder (7B) local_ollama 21.4%
  38. 38 Gemma 4 E2B Ollama/local / local_ollama 21.4%
  39. 39 Phi-4 mini 3.8B Ollama/local / local_ollama 10.7%
  40. 40 Gemma 3 4B Ollama/local / local_ollama 10.7%
  41. 41 Phi-4-Reasoning-plus Ollama/local / local_ollama 7.14%
  42. 42 Gemma 3 270M Ollama/local / local_ollama 3.57%

Source: Official HEPToolBench 1.2 leaderboard

Source: Official HEPToolBench 1.2 leaderboard

#EntryScore ±SettingDateSource
1 Gemini 3.1 Flash-Lite 71.4% 52.9 to 84.7 Google API / api_cloud Mar 2026 Official leaderboard
2 GPT-4.1 67.9% 49.3 to 82.1 GitHub Models API / api_cloud Apr 2025 Official leaderboard
3 gpt-oss-120b 67.9% 49.3 to 82.1 Ollama/local / local_ollama Aug 2025 Official leaderboard
4 GPT-4.1 mini 64.3% 45.8 to 79.3 GitHub Models API / api_cloud Apr 2025 Official leaderboard
5 Llama 3.3 70B (API) open weights, 70B 57.1% 39.1 to 73.5 GitHub Models API / api_cloud Official leaderboard
6 Qwen3-Next-80B-A3B 57.1% 39.1 to 73.5 Ollama/local / local_ollama Sep 2025 Official leaderboard
7 Gemini 3.5 Flash 57.1% 39.1 to 73.5 Google API / api_cloud May 2026 Official leaderboard
8 Magistral Medium 2509 closed weights 53.6% 35.8 to 70.5 Mistral API / api_cloud Official leaderboard
9 Llama 3.3 70B (local) open weights, 70B 50% ±17.4 Ollama/local / local_ollama Official leaderboard
10 Qwen3-14B 50% ±17.4 Ollama/local / local_ollama Apr 2025 Official leaderboard
11 gpt-oss-20b 50% ±17.4 Ollama/local / local_ollama Aug 2025 Official leaderboard
12 Gemma 4 26B A4B 50% ±17.4 Ollama/local / local_ollama Apr 2026 Official leaderboard
13 Gemma 4 31B IT 50% ±17.4 Ollama/local / local_ollama Apr 2026 Official leaderboard
14 DeepSeek-R1 70B open weights, 70B 46.4% 29.5 to 64.2 Ollama/local / local_ollama Official leaderboard
15 Mistral Large 2512 open weights, 675B 46.4% 29.5 to 64.2 Mistral API / api_cloud Official leaderboard
16 Mistral Medium 2508 closed weights 46.4% 29.5 to 64.2 Mistral API / api_cloud Official leaderboard
17 GPT-4.1 nano 46.4% 29.5 to 64.2 GitHub Models API / api_cloud Apr 2025 Official leaderboard
18 Qwen3-Coder-Next 46.4% 29.5 to 64.2 Ollama/local / local_ollama Feb 2026 Official leaderboard
19 Qwen3.5-27B 46.4% 29.5 to 64.2 Ollama/local / local_ollama Feb 2026 Official leaderboard
20 Devstral 2 42.9% 26.5 to 60.9 Mistral API / api_cloud Dec 2025 Official leaderboard
21 Sarvam 105B 42.9% 26.5 to 60.9 Sarvam API / api_cloud Mar 2026 Official leaderboard
22 Gemma 4 12B 42.9% 26.5 to 60.9 Ollama/local / local_ollama Jun 2026 Official leaderboard
23 Codestral 2508 closed weights 39.3% 23.6 to 57.6 Mistral API / api_cloud Official leaderboard
24 Falcon 3 10B open weights, 10B 39.3% 23.6 to 57.6 Ollama/local / local_ollama Official leaderboard
25 Phi-4 14B open weights, 14B 39.3% 23.6 to 57.6 Ollama/local / local_ollama Official leaderboard
26 Gemma 3 12B 39.3% 23.6 to 57.6 Ollama/local / local_ollama Mar 2025 Official leaderboard
27 Devstral 24B open weights, 24B 35.7% 20.7 to 54.2 Ollama/local / local_ollama Official leaderboard
28 Qwen3-8B 35.7% 20.7 to 54.2 Ollama/local / local_ollama Apr 2025 Official leaderboard
29 Qwen2.5-Coder-14B 32.1% 17.9 to 50.7 Ollama/local / local_ollama Sep 2024 Official leaderboard
30 Ministral 3 14B 32.1% 17.9 to 50.7 Ollama/local / local_ollama Dec 2025 Official leaderboard
31 Gemma 4 E4B 32.1% 17.9 to 50.7 Ollama/local / local_ollama Mar 2026 Official leaderboard
32 Granite 4 32B-A9B open weights, 32B 28.6% 15.3 to 47.1 Ollama/local / local_ollama Official leaderboard
33 Mistral Small 3.2 24B open weights, 24B 28.6% 15.3 to 47.1 Ollama/local / local_ollama Official leaderboard
34 Qwen3.5-35B-A3B 28.6% 15.3 to 47.1 Ollama/local / local_ollama Feb 2026 Official leaderboard
35 Llama 3-8B 21.4% 10.2 to 39.5 Ollama/local / local_ollama Apr 2024 Official leaderboard
36 Llama 3.1-8B 21.4% 10.2 to 39.5 Ollama/local / local_ollama Jul 2024 Official leaderboard
37 Qwen2.5-Coder (7B) 21.4% 10.2 to 39.5 local_ollama Sep 2024 Official leaderboard
38 Gemma 4 E2B 21.4% 10.2 to 39.5 Ollama/local / local_ollama Mar 2026 Official leaderboard
39 Phi-4 mini 3.8B open weights, 3.8B 10.7% 3.7 to 27.2 Ollama/local / local_ollama Official leaderboard
40 Gemma 3 4B 10.7% 3.7 to 27.2 Ollama/local / local_ollama Mar 2025 Official leaderboard
41 Phi-4-Reasoning-plus 7.14% 2 to 22.6 Ollama/local / local_ollama Apr 2025 Official leaderboard
42 Gemma 3 270M 3.57% 0.6 to 17.7 Ollama/local / local_ollama Aug 2025 Official leaderboard
#ModelScore ±SettingDateSource
1 Gemini 3.1 Flash-Lite 71.4% 52.9 to 84.7 Google API / api_cloud Mar 2026 Official leaderboard
2 GPT-4.1 67.9% 49.3 to 82.1 GitHub Models API / api_cloud Apr 2025 Official leaderboard
3 gpt-oss-120b 67.9% 49.3 to 82.1 Ollama/local / local_ollama Aug 2025 Official leaderboard
4 GPT-4.1 mini 64.3% 45.8 to 79.3 GitHub Models API / api_cloud Apr 2025 Official leaderboard
5 Qwen3-Next-80B-A3B 57.1% 39.1 to 73.5 Ollama/local / local_ollama Sep 2025 Official leaderboard
6 Gemini 3.5 Flash 57.1% 39.1 to 73.5 Google API / api_cloud May 2026 Official leaderboard
7 Qwen3-14B 50% ±17.4 Ollama/local / local_ollama Apr 2025 Official leaderboard
8 gpt-oss-20b 50% ±17.4 Ollama/local / local_ollama Aug 2025 Official leaderboard
9 Gemma 4 26B A4B 50% ±17.4 Ollama/local / local_ollama Apr 2026 Official leaderboard
10 Gemma 4 31B IT 50% ±17.4 Ollama/local / local_ollama Apr 2026 Official leaderboard
11 GPT-4.1 nano 46.4% 29.5 to 64.2 GitHub Models API / api_cloud Apr 2025 Official leaderboard
12 Qwen3-Coder-Next 46.4% 29.5 to 64.2 Ollama/local / local_ollama Feb 2026 Official leaderboard
13 Qwen3.5-27B 46.4% 29.5 to 64.2 Ollama/local / local_ollama Feb 2026 Official leaderboard
14 Devstral 2 42.9% 26.5 to 60.9 Mistral API / api_cloud Dec 2025 Official leaderboard
15 Sarvam 105B 42.9% 26.5 to 60.9 Sarvam API / api_cloud Mar 2026 Official leaderboard
16 Gemma 4 12B 42.9% 26.5 to 60.9 Ollama/local / local_ollama Jun 2026 Official leaderboard
17 Gemma 3 12B 39.3% 23.6 to 57.6 Ollama/local / local_ollama Mar 2025 Official leaderboard
18 Qwen3-8B 35.7% 20.7 to 54.2 Ollama/local / local_ollama Apr 2025 Official leaderboard
19 Qwen2.5-Coder-14B 32.1% 17.9 to 50.7 Ollama/local / local_ollama Sep 2024 Official leaderboard
20 Ministral 3 14B 32.1% 17.9 to 50.7 Ollama/local / local_ollama Dec 2025 Official leaderboard
21 Gemma 4 E4B 32.1% 17.9 to 50.7 Ollama/local / local_ollama Mar 2026 Official leaderboard
22 Qwen3.5-35B-A3B 28.6% 15.3 to 47.1 Ollama/local / local_ollama Feb 2026 Official leaderboard
23 Llama 3-8B 21.4% 10.2 to 39.5 Ollama/local / local_ollama Apr 2024 Official leaderboard
24 Llama 3.1-8B 21.4% 10.2 to 39.5 Ollama/local / local_ollama Jul 2024 Official leaderboard
25 Qwen2.5-Coder (7B) 21.4% 10.2 to 39.5 local_ollama Sep 2024 Official leaderboard
26 Gemma 4 E2B 21.4% 10.2 to 39.5 Ollama/local / local_ollama Mar 2026 Official leaderboard
27 Gemma 3 4B 10.7% 3.7 to 27.2 Ollama/local / local_ollama Mar 2025 Official leaderboard
28 Phi-4-Reasoning-plus 7.14% 2 to 22.6 Ollama/local / local_ollama Apr 2025 Official leaderboard
29 Gemma 3 270M 3.57% 0.6 to 17.7 Ollama/local / local_ollama Aug 2025 Official leaderboard

Number one over time

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

Google DeepMind
Own leaderboard Not tested since Aug 2026

Source: Official HEPToolBench 1.2 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% 2024 2025 2026 Near the ceiling (85%) Gemini 3.1 Flash-Lite GPT-4.1 Gemma 3 12B Qwen2.5-Coder-14B Llama 3-8B 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Gemini 3.1 Flash-Lite Gemma 3 12B

Source: Official HEPToolBench 1.2 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 (42 entries).
  • Pass Models scored 29 models we track, plus 13 other entries on its leaderboard.
  • Pass Newest score The newest is from June 2026, for Gemma 4 12B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3.5 Flash (#44).
  • Pass Headroom The best is 71.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 29 rows for models we track and 13 other entries 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