DeepSeek Coder V2 Lite Base: benchmark results
Provider: DeepSeek.
Aggregate Rating 1359 ± 29, rank #7461 of 10916 rated models, from 8 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| GSM8K | 67.1 | Accuracy (%) | 68.8 |
| ARC Challenge (AI2) | 57.3 | Accuracy (%) | 52.6 |
| WinoGrande | 72.9 | Accuracy (%) | 45.6 |
| C3-Bench (Controllable Code Completion) - Implementation Control | 27.5 | Instruction-Following Rate (%; 1,286 ICC tasks: passes the u | 42.5 |
| C3-Bench (Controllable Code Completion) - Implementation Control - Pass@1 | 43.7 | Pass@1 (%; unit tests on the 1,286 instructed ICC tasks) | 37.5 |
| MMLU | 60.5 | Accuracy (%) | 36.6 |
| Epoch AI - ECI | 99.89 | ECI Score | 16.4 |
| C3-Bench (Controllable Code Completion) - Scale Control | 4.1 | Instruction-Following Rate (%; 909 SCC tasks: AST node-type | 15 |
Interactive version: theaggregate.ai/model?slug=deepseek-coder-v2-lite-base · How It Works · Data refreshed daily, snapshot 2026-11-08.