DeepSeek officially launches V4.1-Flash-1017 with Pareto gains, wish-style coding and search API
DeepSeek · wechat · 2026-10-17
DeepSeek's official release notes for V4.1-Flash-1017: the new model is live across web, app and API, with broad benchmark gains plus shorter runtimes and outputs — a "Pareto improvement" on performance-cost across multiple harnesses. Optimized for DeepSeek Harness (upgrade to 0.2.1-rc.1+), Tencent CodeBuddy and WorkBuddy.
From-scratch project building: self-feedback RL boosts game-dev and 3D modeling "wish-style" coding, demoed with a Forbidden City 3D scene, a browser racing game, a Godot Portal 2 recreation, and Valley of Ages.
Search API now public: the AI-native engine indexes hundreds of billions of pages, filtering SEO spam, GEO manipulation and low-quality AIGC. New Search and Fetch endpoints cost 5 RMB/1k calls; server-side calls in DeepSeek Harness/Claude Code/Codex are free until Nov 1, 2026.
Open weights: available on HuggingFace and ModelScope as DeepSeek-V4.1-Flash-1017; API name (deepseek-flash) and pricing unchanged.
More from coding & agent
- Palo Alto Networks ships Cortex XCOR: AI SRE agent diagnoses alerts in under 3 minutes — shashib · 2026-10-17
- HF dev begs OpenAI/Anthropic to fix agents picking the wrong Chrome window — NielsRogge · 2026-10-17
- An agent negotiated a 3PL contract in 3 days, end to end — aarthir · 2026-10-17
- Gemini 3 vs Opus 4.5 vs GPT-5.1 vs Grok 4: same-prompt 3D diorama showdown — cedric_chee · 2026-10-17
- Reverse-engineering OpenAI Atlas browser's system prompt with Claude Code — DionysianAgent · 2026-10-17
- Hugging Face Hub adds RL Environment category with 153 agent-ready environments — mervenoyann · 2026-10-17