LLM Stats Score: leaderboard
A composite from llm-stats.com rather than a task a model performs. Published benchmark results and approved community evaluations are compared within each eligible benchmark and then combined across the catalogue, and benchmarks a model was never run on stay missing, not counted as losses. The site labels its index figures TrueSkill conservative ratings, mean minus three standard deviations; higher is better.
Metric: LLM Stats Score (conservative rating). Source: llm-stats.com. Saturation forecast: Around November 2027. 405 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 | 59.93 | #3 |
| 2 | Gemini 4 Carbon | 58.81 | #1 |
| 3 | Claude Sonnet 5.5 | 57.85 | #14 |
| 4 | GPT-6 Astra | 57.32 | #6 |
| 5 | Kimi K3.1 Pro | 55.57 | #17 |
| 6 | Gemini 4 Argon | 54.38 | #9 |
| 7 | Claude Mythos Preview | 54.27 | #5 |
| 8 | Claude Opus 5 | 53.4 | #16 |
| 9 | GPT-6.1 Sol | 53.35 | #7 |
| 10 | Claude Fable 5 | 52.8 | #20 |
| 11 | Claude Fable 5.1 | 52.76 | #10 |
| 12 | GPT-5.6 Sol | 52.68 | #22 |
| 13 | Muse Spark 1.3 | 52.01 | #31 |
| 14 | Kimi K3 | 51.35 | #35 |
| 15 | DeepSeek V4.1 Flash (1017) | 50.88 | #27 |
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-score · How It Works · Data refreshed daily, snapshot 2026-11-05.