ComplexConstraints
Chat & Writing · 2026-06-03
Instruction-following benchmark for interlocking, conditional constraints in professional work.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 57.7 |
| Gemini 4 Argon | 53.3 |
| Muse Spark 1.3 | 51.9 |
| GPT-6 Sol | 50.8 |
| GPT-5.6 Sol | 50.5 |
| Claude Opus 5.5 | 49.9 |
| DeepSeek V4.1 Flash | 49.7 |
| GPT-5.5 | 49.5 |
| Grok 4.7 | 48.9 |
| Gemini 3.8 Flash | 48.4 |
| GPT-5.4 | 46.0 |
| Claude Fable 5.1 | 45.1 |
| Gemini 3.1 Pro Preview | 43.7 |
| Gemini 3.7 Flash | 42.4 |
| Gemini 3.6 Flash | 40.0 |
| DeepSeek V4 Flash Vision Exp | 39.9 |
| Muse Spark 1.2 | 39.9 |
| Claude Fable 5 | 38.1 |
| Muse Spark 1.1 | 38.1 |
| Kimi K3 | 37.9 |
| Claude Opus 5 | 37.3 |
| Grok 4.6 | 36.5 |
| Grok 4.5 | 35.9 |
| Opus 4.8 | 35.6 |
| Hy3 | 34.3 |
| Qwen3.7-Max (source unspecified) | 33.5 |
| GLM-5.3 | 30.5 |
| Kimi K2.6 | 30.1 |
| GLM-5.2 | 29.3 |
| GLM-5.3 Flash | 27.6 |
| Muse Glimmer 30B | 23.9 |
| Gemini 3.5 Flash-Lite | 23.2 |
| Kimi K2.5 | 18.2 |
| Qwen3.5 Plus (2026-02-15) | 17.8 |
| Grok 4.20 | 16.4 |
| ERNIE 5.1 | 14.7 |
| Nemotron 3.5 Lightning 30B A3B (precision unspecified) | 7.1 |
| Mistral Large 3 675B Instruct 2512 | 0.4 |
| ERNIE 4.5 300B A47B | 0.0 |