LLM Stats (COLLIE): leaderboard
Controllable text generation benchmark testing systematic ability to follow constraints on length, format, keyword inclusion, and structural requirements.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 99 | #98 |
| 2 | O3 Mini | 98.7 | #272 |
| 3 | O3 | 98.4 | #124 |
| 4 | Mistral Medium 3.5 | 95.8 | #353 |
| 5 | Seed 2.0 Mini | 91.2 | #276 |
| 6 | GPT-4.5 | 72.3 | #274 |
| 7 | GPT-4.1 | 65.8 | #248 |
| 8 | Mistral Small 4 | 62.9 | #517 |
| 9 | GPT-4o | 61 | #343 |
| 10 | GPT-4.1 Mini | 54.6 | #355 |
| 11 | GPT-4.1 Nano | 42.5 | #731 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-collie · How It Works · Data refreshed daily, snapshot 2026-10-20.