Frontier, p. 5
Frontier
76 stories · page 5 of 9Gemini 3.7 Flash arrives at half the price of the one before it
Google shipped its workhorse model for coding and agents at $0.75 per million input tokens. That is half what 3.6 Flash started at, and it holds until the end of the year.
Read →DeepSeek released V4 Pro with 1.65 trillion parameters and an MIT licence
The model ships with open weights and the most permissive licence anyone hands out. You cannot run it at home, but any provider can stand it up and sell it to you.
Read →Jeff Dean left Google after 27 years and started Discovery Loop
On August 5, Jeff Dean, Sanjay Ghemawat, Quoc Le and Oriol Vinyals left Google and launched Discovery Loop - a public benefit corporation, with a mission to automate research cycles in frontier science. The departure was announced in the same letter from Sundar Pichai that put Koray Kavukcuoglu in charge of Gemini.
Read →Google handed Gemini to Koray Kavukcuoglu - Demis Hassabis moves up to a higher seat
As of August 5, Koray Kavukcuoglu is senior vice president of Google DeepMind and reports directly to Sundar Pichai - he takes over the Gemini models, frontier research, the Gemini app and the developer teams. Demis Hassabis remains Chairman of Google DeepMind and Chief Scientist of Alphabet, and continues to lead Isomorphic Labs.
Read →Claude did not solve the Riemann hypothesis. It raised the bar beneath it
Anthropic published a mathematical result produced with Claude: the proven share of zeta function zeros on the critical line rises from 41.6% to 67.2%. Not a solution to the hypothesis. Something rarer: a verifiable step.
Read →Meta returns to open weights: Muse Glimmer, 30 billion, Apache 2.0
The new model from Meta Superintelligence Labs is multimodal, built for agents and released under the cleanest license there is. Weights on Hugging Face, package on Ollama, zero geographic strings - Europe included.
Read →Qwen released 27 billion under Apache 2.0, with a 2.4 trillion giant next to it
On 3 August we wrote that Alibaba was showing Qwen3.8-Max and promising its weights the following week. The promise has been kept: on 8 August the giant goes up for download, and three days before it, a small 27 billion model under Apache 2.0.
Read →Anthropic touched the guard, not the model - and biological refusals fell by about 85%
Since 7 August, Fable 5 answers health and study questions far more often. The change isn't in the model itself, but in the rules of the classifier that decides what counts as dual-use. Virology, toxicology and molecular design stay closed.
Read →Free ChatGPT moves to a stronger model, with unlimited text chat
From 6 August, paid users get a retuned model with a slider for how deep it thinks. Free users switch to GPT-5.6 Luna by default this week, and get unlimited text conversations from next week. Files and images stay capped.
Read →