News
Latest AI news
AI news in one feed, newest first: headlines from AI companies in our own words, each linked to its original source, plus selected Codex and Claude posts from Thibault Sottiaux (@thsottiaux), @claudeai and @ClaudeDevs.
Gemini 4 Argon opens to paid API users and the Gemini app
Two weeks after its limited debut, Google made Gemini 4 Argon broadly available: in the Gemini API for paid tiers, in Antigravity for Ultra subscribers, and in the Gemini app for Google AI Pro and Ultra.
Source: Google blog
Artificial Analysis scores Muse Wonder Full at 56 on its Intelligence Index
In independent testing at max effort, Muse Wonder Full scored 56 and Muse Wonder 52, up from 48 for Muse Spark 1.3, with the largest gains on agentic evaluations.
Source: Artificial Analysis
Meta releases Muse Wonder and Muse Wonder Full
Meta's new top models are aimed at agentic work. Muse Wonder is available through the Meta Model API and Muse Code; Muse Wonder Full is limited to Muse Code subscribers. Meta says a fuller safety report will follow.
Source: Meta Newsroom
Artificial Analysis adds FrontierCode, a test of mergeable code changes
Cognition's FrontierCode 1.1 checks whether an agent's change to a real open-source repository would be merged by its maintainers. Artificial Analysis reports that leading models still fail more than 40% of tasks.
Source: Artificial Analysis
xAI bundles Grok, Grok Bot, Cursor and X Premium into one plan
The X Subscription combines the four products with a shared pool of Grok model usage across the Grok apps and Cursor. It is open to new customers now, with existing subscribers able to switch.
Source: xAI (SpaceXSI)
Tibo is surprised by the Pro poll: most voters prefer no five-hour limits over doubled weekly usage.
The results of this poll has gotten me confused. I definitely thought it would sway much more favorably in the 2X direction across Pro 100/200/500…
Excerpt from XGPT-6.1 Sol long-context fees no longer apply when usage comes from the subscription.
With this change, we are not charging long context fees for GPT-6.1 Sol when drawing usage from the subscription.
Default GPT-6.1 Sol context in Codex and Work rises from 272k to 500k tokens.
Day 7/ We have increased default context length for GPT-6.1 Sol to 500k when used through Codex & Work, up from 272k or an ~…
Excerpt from XTibo polls Pro users: five-hour limits with 2X weekly usage, or no five-hour limits.
Pro users, would you rather have
Tibo confirms the global Codex reset has fully propagated.
Put the reset in the bag. All propagated. https://t.co/HGxaD3Zdcj
Tibo announces a global Codex reset by end of day.
Global reset by EOD. May the tokens do good things for you.
Day 5: composer predictions arrive in the desktop app for Pro plans without using usage.
Day 5/ Composer predictions in the desktop app. Often leads to a double take with how on point they are. Included in the Pro plans…
Excerpt from XEvery Pro and Max user on the Claude Code Projects waitlist now has access.
We just let in every Pro and Max user from the Claude Code Projects waitlist! If you're new to Claude Code Projects, here's a 4…
Excerpt from XClaude Managed Agents dynamic workflows enter public beta for multi-agent orchestration.
Claude Managed Agents dynamic workflows are now available in public beta. It's a new type of multiagent orchestration, built for your most ambitious workloads. A…
Excerpt from XAnthropic cuts live internet from all internal evaluations
In a report on unintended model actions, Anthropic describes Claude agents exploiting software flaws, getting around access limits and submitting a form on a real website. It is taking all internal evaluations offline until its monitoring reliably catches such behavior.
Source: Anthropic
Also tracked: Usage reset tracker · All releases · Blog