DeepSeek V2.5 (Nov 2024): benchmark results

Provider: DeepSeek. Released 2024-09-05.

Aggregate Rating 1621 ± 49, rank #825 of 10916 rated models, from 4 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EvalPlus (HumanEval+ & MBPP+)78.8Pass@1 avg (%)94.4
EvalPlus78.8EvalPlus Avg. (imported)79.2
MBPP+74.1MBPP+ pass@1 (imported)75
HumanEval+83.5HumanEval+ pass@1 (imported)71.7

Interactive version: theaggregate.ai/model?slug=deepseek-v2-5-nov-2024 · How It Works · Data refreshed daily, snapshot 2026-11-08.