osmosMind
← Leaderboard

STEM Reasoning

Ranked by Accuracy %·8 models·Updated June 2026

#ModelScore
1
OpenAI

GPT-5.4

OpenAI

88.9
2
Anthropic

Claude Opus 4.8

Anthropic

87.6
3
Google

Gemini 3 Pro

Google DeepMind

86.2
4
DeepSeek

DeepSeek V4

DeepSeek

81.0
5
Grok

Grok 4

xAI

79.1
6
AlibabaCloud

Qwen3 Max

Alibaba

78.4
7
Anthropic

Claude Sonnet 4.6

Anthropic

77.9
8
Meta

Llama 4.1 405B

Meta

69.5

Scores are illustrative sample data.

What it measures

STEM Reasoning targets the hard middle of scientific problem-solving: long derivations, counterfactuals, and multi-step tasks where the final answer hides a chain of reasoning. We grade both the answer and the steps.

Methodology

  • Expert-authored itemsSpecialists write from genuine difficulty, not retrieval, and hand over a verifiable reference solution with each item.

  • Step-level rubricsA per-item rubric credits the reasoning chain, so sound-but-partial work counts — not just the final token.

  • Difficulty tiersStratifying and cross-checking each item keeps the distribution controlled and the score interpretable.


← Back to Leaderboard