DeepSeek: DeepSeek V3.1 Terminus

DeepSeek V3.1 Terminus is a text-in, text-out model from DeepSeek with a 163,840-token context window and a 32,768-token output ceiling. It accepts text input only, with no image or audio support noted. Tool use and reasoning are both supported, while structured output availability is unconfirmed based on current data. At $0.27 per million input tokens and $0.95 per million output tokens, pricing sits on the lower end for a reasoning-capable model, which makes it worth evaluating for cost-sensitive workloads involving long documents or multi-step tool calls. Its blended benchmark score of 34.5 comes from a single independent benchmark, so treat that figure as preliminary rather than settled. Teams needing broad performance validation across multiple evaluations should weigh that limited coverage carefully before committing to this model at scale.