Athene-V2-Chat: benchmark results
Nexusflow's RLHF chat fine-tune of Qwen2.5-72B-Instruct, pitched as on par with GPT-4o across chat, math and coding benchmarks (November 2024). Provider: Nexusflow. Released 2024-11-14. Access: Open.
Aggregate Rating 1565 ± 26, rank #1486 of 10916 rated models, from 41 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BigCodeBench | 47.2 | Pass@1 (%) | 89.6 |
| BRIDGE Medical Leaderboard - Zero-Shot | 41.69 | Average Performance (%) | 87.4 |
| BRIDGE Medical Leaderboard - Few-Shot | 50.68 | Average Performance (%) | 86 |
| BRIDGE Medical Leaderboard - CoT | 39.34 | Average Performance (%) | 85.6 |
| U-MATH - Precalculus | 85 | Accuracy (%) | 72.7 |
| U-MATH - Integral Calculus | 31.25 | Accuracy (%) | 69.7 |
| U-MATH - Multivariable Calculus | 55.62 | Accuracy (%) | 69.7 |
| U-MATH | 54.91 | Accuracy (%) | 66.7 |
| U-MATH - Differential Calculus | 36.82 | Accuracy (%) | 66.7 |
| U-MATH - Sequences & Series | 57.79 | Accuracy (%) | 66.7 |
| U-MATH - Algebra | 74.44 | Accuracy (%) | 47 |
| Chatbot Arena (Text - Math) | 1311 | Arena Score | 40.9 |
Interactive version: theaggregate.ai/model?slug=athene-v2-chat · How It Works · Data refreshed daily, snapshot 2026-11-08.