WildBench: leaderboard
Real-user instruction-following benchmark reporting WildBench scores, task-category scores, Elo, and comparison metrics.
Metric: WB Score Task-Macro. Source: github.com. Source: huggingface.co. Saturation forecast: Estimated already saturated. 63 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-05-13) | 59.3 | #325 |
| 2 | GPT-4o Mini (2024-07-18) | 57.14 | #558 |
| 3 | Mistral Large 2 (Jul) | 55.57 | #484 |
| 4 | GPT-4 Turbo | 55.22 | #404 |
| 5 | Claude 3.5 Sonnet (20240620) | 54.7 | #299 |
| 6 | DeepSeek V2 Chat (0628) | 53.99 | #400 |
| 7 | gemma-2-9B-it-SimPO | 53.28 | #580 |
| 8 | gemma-2-9B-it-DPO | 53.22 | #678 |
| 9 | Gemini 1.5 Pro | 52.95 | #445 |
| 10 | GPT-4 Preview (0125) | 52.28 | #486 |
| 11 | Claude 3 Opus (20240229) | 51.71 | #412 |
| 12 | Yi Large | 48.93 | #565 |
| 13 | Gemini 1.5 Flash | 48.85 | #618 |
| 14 | Gemma 2 27B (IT) | 48.54 | #645 |
| 15 | DeepSeek V2 Chat | 48.21 | #563 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=wildbench · How It Works · Data refreshed daily, snapshot 2026-10-20.