DeepSeek Coder V3
SpecializedOpen SourceAPI Available
Benchmark Scores
82.9/100MMLU78.5
HumanEval94
Math72
Reasoning76.5
Coding93.5
About DeepSeek Coder V3
DeepSeek Coder V3 is a specialized coding model that rivals proprietary solutions on code generation benchmarks. It supports 100+ programming languages and excels at code completion, debugging, and explanation.
Strengths
- +Top-tier code generation
- +Excellent code completion
- +Multi-language support
- +Open source
Weaknesses
- -Narrow focus on coding
- -Weaker general knowledge
- -Less conversational ability
Pricing
Per 1M tokens
Input$0.20
Output$0.80
Quick Facts
- Context Window
- 128K
- Parameters
- 236B (MoE)
- Release Date
- 2025-09-15
- Category
- Specialized
Compare Models
See how DeepSeek Coder V3 stacks up against other models