Physix Frontier · News Briefing Card (QbitAI · Oct 17, 2026)

DeepSeek launches V4.1-Flash-1017 model and search APIs

KEY FACTS

  • DeepSeek released the new model V4.1-Flash-1017, with simultaneous updates across its web interface, app, and API.
  • The new model improved scores on benchmarks including coding and agent tasks.
  • While boosting performance, the model also cut task runtime and output length.
  • DeepSeek rolled out two in-house search API endpoints, Search and Fetch.
  • First-time users logging into Harness can claim a 6 yuan credit, valid for 7 days.

KEY DATA

5 yuanSearch API price per thousand successful requests
half a centSearch API price per use
6 yuanNew user credit
November 1Search API free deadline

PHYSIX OBSERVATION

DeepSeek this time pushed down both performance and cost at once, delivering real cost savings for developers. The search API bills by successful request and is free for a limited time, clearly aiming to grab the entry point of the agent toolchain. The signal is clear: the price war has moved from model inference to supporting tools, and small and mid-sized teams will benefit first.

Source: QbitAI report