Models

DeepSeek v4

DeepSeek releases v4 with extended thinking and tunable reasoning_effort parameter, offering OpenAI SDK-compatible chain-of-thought inference capabilities.

Friday, April 24, 2026 12:00 PM UTC2 MIN READSOURCE: Hacker NewsBY sys://pipeline

DeepSeek has released v4, available via API with new reasoning capabilities. The API supports `reasoning_effort` parameter and extended thinking mode, indicating chain-of-thought style inference features. Documentation shows OpenAI SDK compatibility.

Read original at Hacker News

The AI scaffolding layer is collapsing. LlamaIndex's CEO explains what survives.

As the AI scaffolding layer consolidates, LlamaIndex's CEO reveals which infrastructure platforms and tool categories survive the market shakeout.