DeepSeek V4.1 Flash 1017 – API Pricing & Suppliers

DeepSeek-V3.1 Terminus is an replace to DeepSeek V3.1 that maintains the mannequin’s authentic capabilities whereas addressing points reported by customers, together with language consistency and agent capabilities, additional optimizing the mannequin’s efficiency in coding and search brokers. It is a big hybrid reasoning mannequin (671B parameters, 37B lively) that helps each considering and non-thinking modes. It extends the DeepSeek-V3 base with a two-phase long-context coaching course of, reaching as much as 128K tokens, and makes use of FP8 microscaling for environment friendly inference. Users can management the reasoning behaviour with the reasoning enabled boolean. Learn extra in our docs

The mannequin improves software use, code technology, and reasoning effectivity, attaining efficiency similar to DeepSeek-R1 on tough benchmarks whereas responding extra shortly. It helps structured software calling, code brokers, and search brokers, making it appropriate for analysis, coding, and agentic workflows.