DeepSeek launches updated V4 1 Flash 1017 model and public web APIs

DeepSeek released DeepSeek-V4.1-Flash-1017 on October 17, 2026, as an updated checkpoint for its V4.1-Flash model. According to MarkTechPost, the new model checkpoint is accessible via the DeepSeek website, application, and application programming interface. Tech Insider, BigGo Finance, and MarkTechPost all report that DeepSeek simultaneously opened public access to its proprietary Search and Fetch APIs. [1][4][5]

MarkTechPost and Tech Insider report that the updated model scored 77.0 on the DeepSWE v1.1 benchmark for software engineering tasks. DeepSeek claims that the update provides faster task completion, shorter outputs, and higher scores compared to its September V4.1-Flash release. MarkTechPost notes that the company describes the update as a Pareto improvement in performance and cost. [1][4]

XenoSpectrum reports that the updated model outperforms Claude Opus 5 on code-fixing benchmarks while maintaining its original pricing structure. The Decoder adds that the updated checkpoint outperforms larger competing models on agentic tasks and reduces overall costs per task. [2][6]

According to MarkTechPost, the newly opened Search and Fetch APIs cost 0.75 dollars per 1,000 calls. MarkTechPost and Dealroom report that these APIs allow software developers to access a first-party web layer alongside the coding model through a single API key. [1][3]