HELM Capabilities - WildBench — leaderboard
Metric: WB Score. Source: crfm.stanford.edu. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K2 | 86.19 |
| 2 | O3 (2025-04-16) | 86.07 |
| 3 | Gemini 2.5 Pro (Preview 03-25) | 85.67 |
| 4 | GPT-4.1 (2025-04-14) | 85.43 |
| 5 | O4 Mini (2025-04-16) | 85.41 |
| 6 | Grok 3 | 84.94 |
| 7 | GPT-4.1 Mini | 83.8 |
| 8 | Claude Opus 4 (20250514) | 83.3 |
| 9 | DeepSeek V3 | 83.05 |
| 10 | GPT-4o (2024-11-20) | 82.84 |
| 11 | DeepSeek R1 0528 | 82.81 |
| 12 | Qwen3 235B A22B FP8 Throughput | 82.78 |
| 13 | Claude Sonnet 4 (20250514) | 82.47 |
| 14 | Gemini 2.5 Flash (Preview 04-17) | 81.72 |
| 15 | Claude 3.7 Sonnet (20250219) | 81.44 |
Interactive version: theaggregate.ai/benchmark?slug=helm-capabilities-wildbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.