FeynmanBench - Full Amplitude: leaderboard

FeynmanBench (2026) gives a multimodal model a machine-generated Feynman diagram (Standard Model QED, electroweak or QCD, tree level to two loops) with minimal labelling conventions and asks it to recover the full physical content of the diagram. Checkpoint 5: the complete scattering amplitude with Dirac structures, signs and symmetry factors, judged by a three-model LLM panel; pass rate in percent, higher is better.

Metric: CP5 pass rate (%), the complete scattering amplitude with Dirac structures, signs, fermion-loop and symmetry factors, judged by a three-model LLM panel, on FeynmanBench's Feynman-diagram tasks (a rendered diagram plus minimal conventions; Standard Model QED, electroweak and QCD diagrams from tree level to two loops), API models at temperature 0, mean of five runs; four models ran on subsets of 165 to 742 items, the others on about 2,000; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 17 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)7.8#99
2Claude Sonnet 4.63.3#173
3Qwen 3.6 Plus1.7#238
4Claude Opus 4.71.5#85
5Qwen 3.5 Plus1.5#242
6GLM-5V Turbo1.2#360
7Claude Opus 4.61#113
8GPT-5.40.9#135
9Seed 2.0 Pro0.9#222
10Qwen 3.6 Max0.2#193
11Claude Sonnet 4.50#261
12GPT-5.20#185
13Gemini 3 Flash0#175
14Kimi K2.50#266
15GPT-5.10#240

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=feynmanbench-full-amplitude · How It Works · Data refreshed daily, snapshot 2026-11-21.