The new version of DeepSeek-V3 (DeepSeek-V3-0324) uses the same base model as the previous DeepSeek-V3-1226, only the post-training method is improved. The new version of the V3 model draws on the reinforcement learning technology used in the training process of the DeepSeek-R1 model, which greatly improves the performance level on reasoning tasks, and has achieved more than GPT in mathematics and code-related evaluation sets...
Provider: deepseek-ai
Context window: 160K
Pricing: $0.22 input / $0.62 output per 1M tokens
Capabilities: serverless, json_mode, function_calling, structured_output
