KillSwitch-Bench: leaderboard
Coding agents write and interpret programs in KillSwitch, an esoteric language built to be hard for LLMs, working only from its published specification. Five tasks (hello world, a temperature conversion, max, and two program-understanding tasks), each attempted twice with a $5 budget; an attempt passes when its program passes the hidden verification. Each model runs under its vendor's agent (Claude Code, Codex) or OpenCode. Score: pass rate (%) over the ten attempts; the site's own ranking also subtracts small cost and code-size penalties, which are left out here. Higher is better. Source: https://bench.killswitch-lang.org/
Metric: Pass rate (%; attempts whose program passed the hidden verification, every task attempted the same number of times). Source: bench.killswitch-lang.org. Saturation forecast: Around January 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 70 | #12 |
| 2 | GPT-6 Astra | 60 | #6 |
| 3 | Claude Opus 5.5 | 60 | #3 |
| 4 | GPT-6.1 Sol | 60 | #7 |
| 5 | Claude Fable 5.1 | 50 | #9 |
| 6 | Gemini 4 Argon | 50 | #8 |
| 7 | MiMo-V2.6-Pro | 40 | #41 |
| 8 | Muse Spark 1.3 | 40 | #25 |
| 9 | Muse Wonder | 40 | #14 |
| 10 | GPT-5.6 Sol | 30 | #20 |
| 11 | GPT-6 Sol | 30 | #16 |
| 12 | Claude Sonnet 5 | 20 | #56 |
| 13 | DeepSeek V4 Pro (0813) | 20 | #52 |
| 14 | Kimi K3 | 10 | #30 |
| 15 | GLM-5.3 | 10 | #39 |
No result here: #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=killswitch-bench · How It Works · Data refreshed daily, snapshot 2026-10-19.