General

ExploitBench leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on ExploitBench is 100.0% by GPT-6 Astra. 4 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	GPT-6 Astra	100.0%	measured
2	GPT-6.1 Sol	99.7%	measured
3	Claude Mythos 5	78.0%	measured
4	GPT-5.6 Sol	73.5%	measured
5	Claude Mythos Preview	69.0%	measured
6	GPT-6 Sol	61.1%	estimated ± 12.2 pp, low confidence
7	GLM-5.3	54.4%	measured
8	GPT-5.6 Terra	52.9%	measured
9	DeepSeek V4.1 Flash	52.5%	estimated ± 12.2 pp, low confidence
10	MiMo-V2.6-Pro	47.9%	measured
11	GPT-5.6 Luna	33.2%	measured
12	Kimi K3	32.0%	measured
13	MiMo-V2.6-Flash	25.3%	measured
14	GPT-5.5	20.0%	estimated ± 12.2 pp, low confidence
15	GPT-6 Luna	12.5%	estimated ± 12.2 pp, low confidence

All leaderboards

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General