TPS-CalcBench - Unit Correctness: leaderboard

Analytical engineering calculation on hypersonic thermal protection system calculation problems from the 420-item expert-reviewed core set of TPS-CalcBench (2026), answered zero-shot as text. Share of items whose answer unit matches the reference unit after canonicalization, in percent; higher is better.

Metric: Unit correctness rate (%): share of items whose predicted unit matches the reference unit after canonicalization on the 420-item high-confidence core set of TPS-CalcBench (analytical calculation problems in hypersonic thermal protection system engineering), zero-shot text-only prompting; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.287.6#185
2GLM-586.2#269
3GPT-5.185.2#240
4Claude Opus 4.585#160
5DeepSeek V3.181.7#511
6GLM-4.776.4#380
7Seed 1.875.7#252
8Claude Sonnet 4.572.1#261
9Claude Haiku 4.566.2#583
10Gemini 3 Flash (Preview)52.6#152
11MiniMax-M2.524.3#600
12MiniMax-M2.118.8#575

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

Interactive version: theaggregate.ai/benchmark?slug=tps-calcbench-unit-correctness · How It Works · Data refreshed daily, snapshot 2026-11-21.