Model Terminal

DeepSeek V3 (DeepSeek Chat)

DeepSeek V3 is a 671-billion-parameter Mixture-of-Experts large language model with 37B active parameters per token, offering strong performance on chat, coding, and reasoning tasks at competitive inference efficiency. It is available via DeepSeek's own chat interface and API platform, as well as third-party hosts such as OpenRouter. Source: supabase.

Value score
Not scored
Context
tokens
Max output
tokens
Price

Capability radar

Identity

Developer
~deepseek
Openness
Modalities
Release
Knowledge cutoff
Deprecation
API docs

Benchmarks

No benchmark scores yet.

Compare nearby

Pricing

Input / 1M
Output / 1M
Speed