AI News — July 21, 2026

Model releases

• [2026-07-20] Anthropic — Fable 5 subscription terms finalized: permanently included on Max and Team Premium (at 50% of limits), metered for Pro and Team Standard. After a run of week-long extensions that kept Fable 5 temporarily included while Anthropic brought inference capacity online, the company settled the terms in an announcement ahead of the July 20 cutover: Max and Team Premium plans now include Fable 5 permanently, capped at 50% of each plan’s weekly usage limits, while Pro and Team Standard lose included access — those users get a one-time $100 usage credit, after which Fable 5 bills at API rates of $10 / $50 per million input/output tokens. This resolves the watch-list item the last several briefings carried as an open question: the previously rumored, unconfirmed “permanent-50%-for-Max vs credits-for-Pro” split is now the confirmed, live billing structure rather than a placeholder deadline. Anthropic still frames the 50% cap as capacity-driven, not a permanent ceiling. It matters because it ends the six-week “will it be included?” limbo with a concrete two-tier answer — and it draws a sharp line between Max/Team Premium (Fable 5 is now a standing plan feature) and Pro/Team Standard (every session past the $100 credit is a metered line item). (The New Stack, official, prior coverage)

  For Max and Team Premium users the wrinkle to internalize is the 50% cap: Fable 5 is included, but it draws on half the plan’s weekly limit, so a heavy day can exhaust its share and quietly drop you back to other models mid-week — worth knowing where that boundary sits before you lean on it.

  For Software Developers: If you’re on Pro and had wired Fable 5 in as the default in Claude Code, the one-time $100 credit is now a burn-down clock — at $10/$50 per million it covers only so many full sessions — so the concrete move is to demote Fable 5 to the genuinely hard tickets and route routine edits back to an included model before the credit runs out, rather than discovering the switch mid-sprint.

Coding agents

• [2026-07-19] Anthropic — Claude Code 2.1.215 stops auto-running the /verify and /code-review skills and names ripgrep the primary search tool. A small behavior change on the heels of the 2.1.211–2.1.214 permission-fix run: Claude Code no longer invokes the /verify and /code-review skills on its own — you now trigger them explicitly with /verify or /code-review when you want a verification or review pass — and the docs now designate ripgrep as the primary codebase-search tool for faster, more consistent searches. It’s a routine point release, not a security fix, but the skill change is worth noting for anyone who had come to rely on Claude Code auto-running a review step: that step is now opt-in. (official)

  Anywhere Claude Code runs unattended — a CI job or an agent loop that assumed a review pass fired automatically — that pass is now simply absent until someone adds an explicit /code-review step, so it’s worth grepping your automation for that assumption.

  For Software Developers: If your habit was to let Claude Code wrap up a task and trust it had self-reviewed, that check is gone — the concrete adjustment is to end each task by running /verify or /code-review yourself, or bake those into a task-completion checklist, so the review you were getting implicitly becomes an explicit last step you actually watch run.

Watch list

• Kimi K3 open weights (July 27) — Moonshot has K3 live via API (kimi-k3, $3/$15 per million) but says the 2.8T weights go public July 27, reportedly under a modified-MIT-style license (unconfirmed until the model card ships). Watch for the model card, the license’s rumored monthly-active-user clause, whether third-party inference providers stand up endpoints undercutting Moonshot’s own pricing, and the first independent, non-arena evals. (prior coverage)

  Independent, non-arena evals landing after the weights drop are the signal worth waiting for here — an arena ranking and a vendor’s own benchmarks are the two least reliable ways to judge whether K3 is real competition, so the July 27 weights matter less as a self-host option than as the moment third parties can finally test it at their own cost.

• Gemini 3.5 Pro — still undated; an August window is now rumored — after the third missed target (July 17) and Bloomberg’s “months behind schedule” report, Google has not published a replacement date, and reporting now floats an August window with an upgraded Flash model in partner testing that could ship first. The headline specs (2M-token context, Deep Think, pricing) remain reported, never confirmed, with no model card. The signal to watch is a Google post with an actual date and a published model card, not another leaked target. (unconfirmed) (prior coverage)

  An upgraded Flash shipping before any Pro date would itself be the first concrete evidence that Google’s pipeline is unblocked at all — a more informative near-term tell than a fourth rumored Pro window, so that release is the thing to watch even more than the date.

• MCP spec finalization (July 28) — the 2026-07-28 release candidate (stateless core, Extensions framework, Tasks, MCP Apps, authorization hardening) is in its validation window, with the Python/TypeScript/Go/C# SDKs updating against it in beta. The Streamable HTTP transport requires the Mcp-Method/Mcp-Name routing headers that security analysts have flagged as fresh attack surface. Watch for the remaining SDK support landing and any last header changes before the RC locks. (official, prior coverage)

  Watch which of the four SDKs (Python, TypeScript, Go, C#) reaches parity last: if the one your stack depends on is still in beta when the RC locks on July 28, you’ll be validating against a moving target while everyone else tests on the final transport.