DeepSeek V3.1 Terminus

DeepSeek released V3.1 Terminus, a 685B-parameter Mixture-of-Experts model with ~37B active parameters per token, featuring improved language consistency and tool-use capabilities including Code Agent and Search Agent. The model is available under an MIT license with FP8 quantization.

DeepSeek V3.1 Terminus MoE workstation685B total, ~37B active per token MoE - same backbone as DeepSeek V3. Instruct/agent-tuned variant of V3.1-Base: improves language consistency less mixed Chinese-English and adds Code Agent + Search Agent tool-use. FP8 quantization. Open weights under MIT. coding reasoning chat - 685.0B - 128k - mit - Sep 2025 Scores Coding 80 Reasoning 85 General 82 PRICE HISTORY Inference cost over time Data accumulates from the first daily sync - longer ranges populate over time. Prices come from OpenRouter snapshots, not a historical API. Loading price history...