Gemini 4 Hydrogen: Flash-tier price, Argon-level smarts
Google's fast new model edges out Gemini 4 Argon on an intelligence index and runs 2.5x faster than 3.8 Flash, at $1.50/M input tokens.
Model releases, benchmarks and capabilities. 131 stories, page 1 of 11.
Large language models are neural networks trained on vast amounts of text, code and increasingly images and audio, then used to answer questions, write, reason and act as the engine behind AI agents. This section follows the releases that move the field: new frontier and open-weight models, benchmark results and what they do and do not measure, context windows, reasoning and tool use.
The practical question behind most LLM news is whether a model is worth building on, so we pair releases with price per million tokens and capability data from our model tracker. Where a story leans on a headline number, we link the primary source so you can see how it was measured.
Google's fast new model edges out Gemini 4 Argon on an intelligence index and runs 2.5x faster than 3.8 Flash, at $1.50/M input tokens.
Google's full Gemini app is now on CarPlay: tap the icon, talk to Gemini Live hands-free, and the chats sync back to your iPhone.
Anthropic's first Fable 5.2 showcase: a 3D cell built in code, origami with printable folds, and a landmark 3D-printed from 40 photos.
Google's top Gemini model is out to paying developers and AI Ultra users at Argon's price. The price doubles on 1 December.
Anthropic adds shared Sheets and Whiteboards (beta) to paid plans, makes Claude Data GA, and lets your apps open right inside Claude.
Anthropic's Fable 5.2 is open to everyone at $8/$40 per million tokens. Mythos 5.2, the same model with looser limits, stays with vetted orgs.
Anthropic's Fable 5.2 is 1 on Artificial Analysis' index at 65, but at max effort a task costs about $26.50, double what Fable 5.1 cost.
For one week, Pro and standard Team seats can spend up to 50% of their weekly Claude limits on Fable 5.2. After that, Fable goes back to usage credits.
Anthropic's new Claude pair is 30% faster and 20% cheaper. Fable 5.2 is open to everyone; Mythos 5.2 stays with vetted orgs.
Google's new video model makes and edits 15-second clips with sound for $0.07 a second at 720p. That's 30% less than Omni 1.1.
Moonshot put its biggest model on Hugging Face. It nearly matches Claude Opus 5.5 on Terminal-Bench, with its hacking skills deliberately cut back.
Anthropic's revised S-1 says internal estimates show it may soon hit its own threshold for AI-accelerated AI progress. No model has crossed it yet.