AI News — June 8, 2026

Model releases

• [2026-06-08] Apple / Google — WWDC 2026 opens today (keynote 10:00 PT / 17:00 UTC), and the centerpiece is the long-delayed, Gemini-powered Siri. The underlying deal is already confirmed: Apple and Google announced in January that the next generation of Apple Foundation Models and Siri’s planner/summarizer would run on a custom ~1.2-trillion-parameter Gemini model (an ~8× jump from the ~150B model behind today’s Apple Intelligence), licensed for a reported ~$1B/year, with Siri requests routed through Apple’s Private Cloud Compute so they’re anonymized and not used to train Google’s models. Today is the public unveiling — the Siri app, multi-step commands, and rollout timing are what’s actually new on stage. (source, source, source)

  The notable part isn’t a benchmark — it’s the deployment. A frontier model reaching a billion-device consumer surface through a third-party license is the largest single distribution event for any model this year, and it’s an Apple admission that its own foundation models weren’t ready to run the assistant. For anyone building on Gemini, it’s also a durability signal: the model now has Apple’s scale behind its roadmap.

  Routing requests through Private Cloud Compute — so Google receives only anonymized traffic it can’t train on — quietly sets a template: it’s the bar other vendors will be measured against when they want to put a rival’s frontier model on a sensitive consumer surface without handing over the underlying data.

• [2026-06-08] Apple — reportedly pairing the Gemini-Siri overhaul with an Apple Intelligence “Extensions” model-choice system that would let users pick a third-party model — ChatGPT, Gemini, or Anthropic’s Claude — to back Apple Intelligence features, the way ChatGPT integration works today. Apple has not confirmed the specifics ahead of the keynote. If it ships as described, Claude and GPT join Gemini as swappable backends in front of the install base rather than one bundled default. (source) (unconfirmed)

  Should it ship, it would recast Apple Intelligence as a routing layer over competing vendors rather than a product with a single model behind it — making “the default backend on a billion devices” a contested slot the way the browser’s default search engine is, and giving users a model preference where today they mostly get whatever Apple bundled.

Watch list

• WWDC 2026 keynote (today, 10:00 PT): the things to actually watch when it airs — whether the Siri app, multi-step commands, and Dynamic Island integration land as previewed; whether the model-choice Extensions (and Claude as an option) are confirmed; and the rollout date, since past Siri AI features slipped repeatedly. Reported to be Tim Cook’s last keynote. (source)

• FinOps X (June 8–10, San Diego) — now live: the Tokenomics Foundation’s technical roadmap, initial working groups, and the FOCUS token-spend extension are due to be detailed at the conference that opens today — the moment last week’s “intent to launch” turns into something teams can adopt for normalizing token spend alongside cloud cost. (source, source)

• Gemini 3.5 Pro (this month): still unshipped — promised “next month” at I/O (May 19), reportedly a 2M-token context window, a “Deep Think” reasoning mode, and pricing around ~$15/$60 per 1M tokens. Watch the model card, API name, and that price on day one. (source) (unconfirmed)

• MiniMax M3 weights (~June 11): M3 launched June 1 with frontier coding claims (59.0% SWE-Bench Pro) but no weights or technical report; MiniMax committed to publishing both on Hugging Face/GitHub “within 10 days,” so the real open-weights test and independent benchmarks land around June 11. (source) (unconfirmed)

• Anthropic developer billing split (June 15): Claude Agent SDK, claude -p, Claude Code GitHub Actions, and third-party agents reportedly move off subscription limits onto a separate monthly credit pool (~$20 Pro / $100 Max 5x / $200 Max 20x), metered at full API rates with no rollover; Anthropic still hasn’t detailed it officially. The sharpest impact is on bursty automation — per-commit CI and GitHub Actions — where full-rate metering turns a usage spike into a bill spike. (source) (unconfirmed)

• MCP spec 2026-07-28 final: stateless core, Tasks, and MCP Apps move to stable, and clients must validate the iss parameter per RFC 9207. Server operators relying on sticky sessions will need to migrate — worth auditing deployments (and iss handling on clients) before the date, especially after the recent VIPER-MCP server-side CVE wave. (source)