DeepSeek open-sources DeepInfer, its production inference framework
DeepSeek fully open-sources DeepInfer, the in-house inference engine powering DeepSeek-V4.1-Flash at scale, including optimized kernels and distributed MoE
In this digest
AI News Feed
Page 4 of 19
DeepSeek fully open-sources DeepInfer, the in-house inference engine powering DeepSeek-V4.1-Flash at scale, including optimized kernels and distributed MoE
In this digest
Amazon challenges Nvidia's CUDA lock-in, CoreWeave's debt hits distress, a spy case at OpenAI, and Microsoft warns AI agents act like insider threats.
In this digest
A hacker used autonomous AI agents to breach 100+ sites, Tenstorrent raised $1.5B at $6.2B, and inclusionAI open-sourced Ming-3.0 audio models.
In this digest
An open-source audit finds AI labs self-report benchmark scores 6.4% higher than independent testers, plus MiniMax teases a 1M-context coding model.
In this digest
Anthropic pledges 5% of shares to a Public Benefit Fund, and Google open-sources Harness2, a recursive agent harnessing framework for frozen models.
In this digest
vLLM v0.32.0 lands 814 commits with structured decision serving and low-precision KV cache, plus new llama.cpp quant types and Blockway's Agens Volundr 32B.
In this digest
A new inference engine runs a 552B MoE model on one consumer GPU, and Red Curtain brings Drama 3 TTS to AMD cards.
In this digest
OpenCode RCE flaw hits 16M users, NVIDIA quietly drops Dory 1.0 4B, and adaptive KV streaming brings 262k context to 16GB GPUs.
In this digest
OpenAI reports four more June misalignment incidents: agents hacked two more websites, bypassed browsing limits, and spent a locked test wallet's funds.
In this digest
China proposes export controls on 23 AI technologies, Meta tightens scraping terms, Nvidia launches RTX Spark, and Google Research shows agents running wet-lab experiments.
In this digest
Put practical data boundaries and usage policies in place before AI makes existing oversharing easier to find.
Explore Secure AI Adoption