Google releases Gemini 4 Neon at $3/$15 per million tokens
Neon has a 1 million-token context window and is rolling out in AI Studio, Antigravity and the Gemini app; the benchmark claims are Google's.
Models, from Google DeepMind
Most of today's numbers come from the companies that produced them: Google on Neon's benchmarks, Microsoft on price against performance, RadixArk on its tuning gains, OpenAI on chip design. The items with immediate work attached are plainer, including two critical NGINX flaws awaiting patches and agent tooling still gated behind a preview.
7 stories, about 9 minutes to read them all.
Neon has a 1 million-token context window and is rolling out in AI Studio, Antigravity and the Gemini app; the benchmark claims are Google's.
Models, from Google DeepMind
Microsoft says MAI Code 2 reaches 98% of GPT-6.1 Sol's performance at about 40% of the price, with Copilot pricing promotional until May 17, 2027.
Models, from microsoft.ai
Both bugs sit in HTTP/2 handling and score 9.2, Wiz says; one reaches back to NGINX 1.9.5 and works in default HTTP/2 setups.
Safety and security, from wiz.io
Checkpoints come in BF16, FP8 and NVFP4; RadixArk says Terminal-Bench 4.1 pass@1 rose from 5.6% to 12.0%.
Open source, from Hugging Face
The post pairs Transformers with expert streaming and a demo Space, but the text available gives no speed figures or setup details.
Open source, from Hugging Face
The PPA gain took under five months, and OpenAI says AMD has put some of the changes into a design sent to manufacturing.
Chips and compute, from The Decoder
The new gh event listen command replaces polling loops, but it is in preview and needs access to Event Streams, which has a waitlist.
Coding tools, from GitHub Changelog