ObviousBench
General QA · 2026-07-16
ObviousBench measures whether language models avoid simple, user-visible mistakes in literal counting, spelling transforms, ordering, negation, formatting, arithmetic, word counting, and basic constraint awareness. Its headline answer pass^3 metric is the percentage of 144 private held-out items for which all three sampled answers are correct.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 100.0 |
| Claude Sonnet 5.5 | 100.0 |
| Claude Opus 5 | 100.0 |
| Gemini 3.6 Flash | 100.0 |
| Gemini 3.7 Flash | 100.0 |
| Gemini 3.8 Flash | 100.0 |
| Gemini 3.1 Pro Preview | 100.0 |
| Gemini 3.5 Flash | 100.0 |
| Gemma 4 31B IT | 100.0 |
| Kimi K3 | 100.0 |
| GPT-5 | 100.0 |
| GPT-5.5 | 100.0 |
| GPT-6 Astra | 100.0 |
| GPT-6 Luna Pro | 100.0 |
| GPT-6 Sol Pro | 100.0 |
| GPT-6.1 Sol | 100.0 |
| o3 | 100.0 |
| Qwen3.5-27B | 100.0 |
| Hy3 | 100.0 |
| Grok 4.5 | 100.0 |
| Grok 4.6 | 100.0 |
| Claude Fable 5 | 99.3 |
| Claude Fable 5.1 | 99.3 |
| Opus 4.5 | 99.3 |
| Opus 4.8 | 99.3 |
| GLM-5.2 | 99.3 |
| Gemini 3 Flash Preview | 99.3 |
| GPT-5.3-Codex | 99.3 |
| Muse Spark 1.3 | 99.3 |
| GPT-6 Luna | 99.3 |
| GPT-6 Sol | 99.3 |
| Qwen3.5 397B A17B | 99.3 |
| Grok 4.7 | 99.3 |
| Grok Build 0.1 | 99.3 |
| Opus 4 | 98.6 |
| Opus 4.1 | 98.6 |
| GLM-5.3 | 98.6 |
| Muse Glimmer 30B | 98.6 |
| Muse Spark 1.1 | 98.6 |
| Muse Spark 1.2 | 98.6 |