DeepSeek previews V4 with 1M context, open weights and agentic coding upgrades
eyishazyer · x · 2026-07-29
DeepSeek says DeepSeek-V4 Preview is officially live and open-sourced, with 1M context now the default across its services.
The release includes two variants:
- DeepSeek-V4-Pro: 1.6T total / 49B active parameters
- DeepSeek-V4-Flash: 284B total / 13B active parameters
DeepSeek claims V4-Pro rivals top closed models, leads open models in world knowledge, and is strong in math, STEM, and coding. V4-Flash is positioned as a faster, cheaper option with similar reasoning on simpler agent tasks.
The company also highlights architectural changes such as token-wise compression plus DSA (DeepSeek Sparse Attention), and says the model is optimized for agentic coding and already integrated with tools like Claude Code, OpenClaw, and OpenCode.
API access is available immediately, and the models support OpenAI ChatCompletions and Anthropic-style APIs.
More from coding & agent
- Building a portable context vault in Notion so every AI agent knows you from day one — ivanhzhao · 2026-10-16
- Reflection AI open-sources Beam: 501B-param MoE model under Apache 2.0 for agentic coding — arena · 2026-10-16
- Deep Research API showdown: OpenAI, Perplexity, Parallel, Tavily, Gemini, Valyu compared — bar_raiser333 · 2026-10-16
- Geoffrey Huntley open-sources Overalls, a self-hosted AI developer platform under MIT — kieranklaassen · 2026-10-16
- Dev shows 199.7bn tokens over 200 days on OpenAI Codex, peak 8.6bn/day — Dimillian · 2026-10-16
- Ask Claude to open its changes in nvim and scroll to the exact line for review — max__drake · 2026-10-16