[REVISION HISTORY]

DeepSeek V4.1-Flash-1017 and DeepInfer release

Updated 2 times since CLSTR started tracking revisions of this situation.

What changed

2026-10-19 03:42 UTC → 2026-10-22 05:21 UTC · added ~~removed~~

DeepSeek V4.1-Flash-1017 ~~AI model~~ and DeepInfer release

DeepSeek has released V4.1-Flash-1017, an updated version of its Flash AI model. The company characterizes the release as a “Pareto improvement” in both performance and cost, noting that the model is faster and produces shorter outputs than previous versions. Technical specifications indicate the model is a sparse mixture-of-experts model featuring a 552B-parameter backbone. It shows improvements in reasoning, coding, agentic tasks, and visual benchmarks. Specifically, the model scored 77.0 on the DeepSWE v1.1 coding benchmark, up from 74.2 in the predecessor. The update utilizes reinforcement learning with self-feedback (RLSF) and employs compressed KV caching to achieve lower costs and faster execution. The model's enhanced capabilities extend to agentic tasks such as 3D modeling and game development. Alongside the model, DeepSeek has opened its Search and Fetch APIs to all users, allowing AI agents and RAG applications to access live web content. The Search API provides structured data, while the Fetch API returns content in Markdown format; both are priced at $0.75 per 1,000 successful requests. The model weights have been released under the MIT license on Hugging Face and ModelScope, a move that intensifies competition between China’s open-weight models and proprietary models from US-based companies. DeepSeek has further expanded its ecosystem with the release of DeepInfer, an open-source software infrastructure designed to run AI models at lower costs. DeepInfer supports the V4.1-Flash model and is compatible with both Nvidia Blackwell GPUs and Huawei Ascend 950 chips. By utilizing distributed inference techniques, the system aims to optimize resource consumption and reduce energy use.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.