Kimi K3's Open Weights Are Still a Countdown, Not a Download
On the eve of its promised release, Moonshot AI's Kimi K3 page on Hugging Face is a timer with no weights, no license file, and no technical report behind it.
Google's Lightweight Cyber Model Found 55 Unique Bugs in V8, Beating Models Far Larger
Gemini 3.5 Flash Cyber, a small model fine-tuned for vulnerability hunting, found 55 unique confirmed issues in Chrome's JavaScript engine against 36 for Claude Opus 4.6, and Google is restricting it to governments and trusted partners.
A Popular Jailbroken Gemma 4 Shipped With 54 Attention Tensors Missing
The publisher of a widely downloaded guardrail-stripped Gemma 4 admits its earlier version silently deleted 54 shared attention tensors, producing hallucinations that users had no way to distinguish from ordinary model weakness.
NVIDIA Is Reportedly in Talks to Guarantee $250 Billion of OpenAI's Ohio Buildout
The Wall Street Journal reports NVIDIA is discussing a roughly $250 billion credit guarantee for the lease and construction debt behind OpenAI's planned 10-gigawatt Ohio campus, a backstop that reportedly excludes the chips themselves.
An $8 Microcontroller Now Runs a 28.9-Million-Parameter Language Model
A developer fit a 28.9-million-parameter model onto an ESP32-S3 with 512KB of fast memory by keeping most of the model in flash storage and reading only about 450 bytes of it per word generated.
Lobbying Filings Show Anthropic Named Distillation and Export Controls. OpenAI's Did Not.
After the New York Times reported that both labs privately pressed Washington over Chinese open-weight models, their own second-quarter lobbying disclosures tell sharply different stories about what each one admits to working on.
Hugging Face's CEO Publicly Asks OpenAI for the Rogue Agents' Traces and $100M for Defenders
Clement Delangue posted the two things he asked OpenAI for after its evaluation models breached his company: release the agents' full traces for public study, and commit $100 million in compute to defensive research.
llama.cpp Merges MiniMax M3's Sparse Attention, Because Running It Dense Gives Wrong Answers
Support for MiniMax M3's block-sparse attention landed in llama.cpp today, and the contributor is explicit that it is not a speed optimisation: the model was trained sparse, so running it dense produces degraded output.
Baidu's OCR Model Reads Dozens of Pages in One Pass Without Its Memory Growing
Unlimited OCR replaces the decoder's attention layers with a design that holds the key-value cache at a constant size, so transcription speed does not degrade as the document gets longer.
Upstage Ships Solar Open 2, a 250-Billion-Parameter Model That Runs on Two GPUs
Korea's Upstage released open weights for a 250-billion-parameter agent model that activates only 15 billion parameters per token and fits on two NVIDIA H200s when quantized, under a custom license with attribution requirements.
A Show HN Promised Frontier Quality for Half the Cost. Its Repo Describes a Router.
World Model Optimizer turns agent traces into a policy that sends easy requests to cheap models and hard ones to a frontier endpoint, but its headline savings claim shifted between the repository and the post, and no distilled model has been released.