DeepSeek
DeepSeek's open V4 model family, the DeepSeek API, and its agent-facing Harness plus Search and Fetch APIs. Every brief item about DeepSeek, newest first, each with its primary source.
Where it stands
DeepSeek is now selling the model together with what agents need around it. On October 17 it released V4.1-Flash-1017 under MIT, a minor update to V4.1-Flash that lifts DeepSWE v1.1 from 74.2 to 77.0 and Terminal-Bench 4.0 from 31.2 to 42.3, and says it also finishes tasks faster with shorter outputs. The same day it opened Search and Fetch APIs built on its in-house search engine to everyone, at $0.75 per 1,000 calls per endpoint.
The trend is clear: beyond models, it is laying down a full agent chain, with its own DeepSeek Harness, server-side tool support in Claude Code and Codex, tuning for Tencent CodeBuddy and WorkBuddy, and now web search. Making search free through other vendors' tools until October 31 is plainly a bid to become developers' default.
What to watch next: real usage and reception of Search and Fetch once billing starts on October 31; whether independent evaluations reproduce the coding scores in its table; whether the larger V4 Pro gets a similar update.
My advice: treat V4.1-Flash-1017 as a low-cost candidate for coding agents and run it head to head with your current model on your own repositories; test search hit rate and latency during the free period, estimate cost from calls per task before adopting it, and keep an interface that lets you switch to another search provider.
Darius · Updated Oct 18, 2026
Timeline 2 items
-
Oct 18 · Sun · 2 items
- Model release huggingface.co ↗DeepSeek releases V4.1-Flash-1017 under MIT: DeepSWE v1.1 rises from 74.2 to 77.0 and Terminal-Bench 4.0 from 31.2 to 42.3
Take · Put it next to Beam on 10-17 and Muse Spark Open on 10-16 and open coding models closed in on closed ones all week. I'd point Claude Code at it on a batch of real tasks from my own repos, checking whether shorter outputs still mean complete changes; note the missing Jinja template before self-hosting, and don't force a generic one onto it.
From the Oct 18 brief · item 01 → - Tools api-docs.deepseek.com ↗DeepSeek opens its Search and Fetch APIs to everyone at $0.75 per 1,000 calls per endpoint, free through Claude Code and Codex until October 31
Take · Agent web access now has a plainly priced option from a Chinese provider. For a content backend that looks things up, like PandaClaws, use the free window before October 31 to compare hit rate and latency against your current search; before adopting it, count searches per task, because that number runs away easily in agents.
From the Oct 18 brief · item 02 →
-