DeepSeek-Coder-V2-Base
DeepSeek · 2024-06-14 · 235.7B parameters
DeepSeek-Coder-V2-Base is the uninstructed 236B-class MoE base checkpoint of DeepSeek-Coder-V2, released June 14, 2024; its downloadable weights contain exactly 235,741,434,880 parameters, while DeepSeek reports rounded figures of 236B total and 21B activated. It was continued-pretrained from an intermediate DeepSeek-V2 checkpoint on 6T additional code-heavy tokens and supports 338 programming languages; DeepSeek documents 128K context although the YaRN configuration declares 163,840 positions.
Benchmark scores
| Benchmark | Score |
|---|
No benchmark scores recorded.