SpaceXSI, the company formerly known as xAI, released Grok Voice 3.0, the speech-to-speech model behind its Voice Agent API, aimed at businesses running customer-support and sales calls. Independent tester Artificial Analysis ranks it first in its category.
Fish Audio's Drama 3 is a 35B mixture-of-experts text-to-speech model with around 3B active parameters that accepts plain-language direction and supports 83 languages. Its weights are published under a community license capped by revenue.
Simon Willison shares Tree Appreciator, a three.js game built by prompting Claude Opus 5.5 in the standard Claude iPhone app rather than Claude Code. The game shows two algorithmically generated trees and asks players to appreciate how they differ.
The Verge argues that the frontier race has reopened after a week in which Anthropic and OpenAI seemed to have pulled away with their latest models. Google had looked far behind and Meta had yet to prove itself despite heavy spending, but recent developments have changed that picture.
OpenAI is racing to finish its next flagship model, which uses the same looped transformer technique as Astra to boost math and coding, though looping makes safety monitoring harder. Employees say one or two new models will likely ship by year's end whether or not the flagship is ready, and Microsoft accidentally revealed details of the technique on a public page.
Subquadratic has a decision model it says offers a context window twelve times larger than Jev's. Jev, the decision model TypeSafe unveiled in mid-September, has prompted a wave of rivals aiming to take work away from frontier language models.