Back
AI summary
Written by AI from the official notes. Check them for exact details.The deepseek-chat model has been upgraded to DeepSeek-V2-0628 with improved reasoning and role-playing capabilities.
- Improved coding performance: HumanEval Pass@1 increased to 84.76%
- Mathematics accuracy improved to 71.02%
- Reasoning capabilities enhanced to 83.40%
- Win rate against GPT-4-0314 increased to 68.3%
- Enhanced role-playing capabilities for character interactions.
Why it matters: Users seeking advanced reasoning and character interaction in conversations should consider this upgrade.
Full release notes3 changes
The deepseek-chat model has been upgraded to DeepSeek-V2-0628.
Model's reasoning capabilities have improved, as shown in relevant benchmarks:
- Coding: HumanEval Pass@1 79.88% -> 84.76%
- Mathematics: MATH ACC@1 55.02% -> 71.02%
- Reasoning: BBH 78.56% -> 83.40%
In the Arena-Hard evaluation, the win rate against GPT-4-0314 increased from 41.6% to 68.3%.
The model's role-playing capabilities have significantly enhanced, allowing it to act as different characters as requested during conversations.