DeepSeek V3 (1226): benchmark results
Provider: DeepSeek. Released 2024-12-26.
Aggregate Rating 1616 ± 43, rank #807 of 10556 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GeoCode Leaderboard | 71.67 | AutoGEEval++ Pass@1 (imported) | 95.5 |
| MARS-Bench - Context Retrieval | 37.08 | Score (0-100) | 41.7 |
| MARS-Bench | 33.16 | Score (0-100) | 33.3 |
| MARS-Bench - Information Reasoning | 28.71 | Score (0-100) | 33.3 |
| MARS-Bench - Task Switching | 29.63 | Score (0-100) | 33.3 |
| MARS-Bench - Instruction Following | 37.23 | Score (0-100) | 16.7 |
Interactive version: theaggregate.ai/model?slug=deepseek-v3-1226 · How It Works · Data refreshed daily, snapshot 2026-11-17.