EN 中
Home
AI News ▾

Explore

AI Articles Track frontier AI news, trends, and industry shifts. AI Daily A quick daily digest of the latest AI hotspots.
AI Products ▾

Explore

Discover Products Find useful AI products and apps for different needs.

Tools

AI Tools Directory A one-stop directory of common AI tools and workflows.
Model Hub ▾

Explore

Model Library Browse AI models across different tasks and scenarios. Model Rankings Compare trending models by capability and popularity. Providers Find model vendors and related AI services.

Tools

Model Compare Compare models side by side for faster selection. Cost Calculator Estimate model usage cost and plan your budget. PC Compatibility Check Check whether your computer can run AI models smoothly. Server Config Calculator Get recommended server specs based on compute demand.
MCP ▾

Information

MCP Servers Discover popular AI-MCP services and find your best match fast. MCP Client Easy MCP client integration and access to powerful AI capabilities. MCP Case Tutorials Master MCP usage from beginner to expert with practical tutorials. MCP Ranking Top MCP service performance rankings to help you choose well.
AI Guides

DeepSeek V4 Lands Under MIT License, V4 Flash Turns Heads for Coding

Published: September 5, 2026 · Updated: October 10, 2026 Read time: 2 min
DeepSeek V4 Lands Under MIT License, V4 Flash Turns Heads for Coding

DeepSeek has released V4, its newest open-weight model family, under the permissive MIT license — with two very large tiers plus a V4 Flash variant that early users compare favorably to closed-lab coding models.

DeepSeek V4 Lands Under MIT License, V4 Flash Turns Heads for Coding

Update (October 10, 2026): The family moved again on September 10: DeepSeek V4.1 Flash shipped under MIT with another 70% output-price cut and now tops BenchLM's open-weight list — making "newest generation" below a snapshot of the August picture rather than the current one.

DeepSeek has released V4, the newest generation of its open-weight models, under the MIT license — the most permissive of the common open-source licenses, imposing essentially no restrictions on commercial use, modification, or redistribution. The family ships in two very large tiers, plus a V4 Flash variant optimized for speed.

Early community reception has centered on coding. In developer discussions, V4 Flash is drawing comparisons to closed-lab fast models — including favorable comparisons against Google's Flash-tier offerings — for real-world software engineering work. Independent benchmark tables are still filling in, but the pattern from DeepSeek's previous releases looks to be holding: capability per parameter that consistently outpaces expectations.

The MIT licensing is the strategically important detail. V4 joins an open-weight cohort — Qwen3.8 under Apache 2.0, GLM-5.3, Kimi K3 — that enterprises can self-host, fine-tune, and embed in products without negotiating terms with an API vendor. For regulated industries and cost-sensitive startups alike, that optionality has become a core part of AI infrastructure planning.

The competitive effect on closed labs is already visible. OpenAI and Anthropic both introduced cheaper tiers this month amid customer pressure on AI spending, and Google's promotional Gemini 3.8 Flash pricing reads partly as a defensive move against exactly this class of open alternative.

Hardware requirements remain the catch: the full V4 tiers are very large, and running them at production scale demands serious GPU infrastructure. The Flash variant narrows that gap. For teams that can meet the infrastructure bar, though, V4 strengthens the case that the open-weight frontier is no longer a generation behind — it is here.