The week's models got safer, cheaper, and closer to your existing stack
Claude Fable 5, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
Anthropic
Claude Fable 5
DEPLOYif you run long-horizon coding or knowledge work
WHAT SHIPPEDAnthropic launched Claude Fable 5, a Mythos-class model made safe for general use. It is state-of-the-art on nearly all tested benchmarks, with the longest and most complex tasks showing the largest lead over prior Claude models. Pricing is $10 per million input tokens and $50 per million output tokens, less than half the price of Claude Mythos Preview.
THE CATCHSafeguards are tuned conservatively and trigger on average in less than 5% of sessions, routing some queries to Claude Opus 4.8 instead. Anthropic does not support Zero Data Retention for Fable 5; prompts and completions are retained for 30 days.
IF YOU RUNa 50-million-line Ruby codebase, Stripe reported Fable 5 completed a codebase-wide migration in a day that would have taken a team over two months
SOURCEAnthropic, Claude Fable 5 and Claude Mythos 5
The Deploy Log
Models and APIs
| DEPLOYif you need multimodal models on laptops Gemma 4 12B. Gemma 4 12B is an encoder-free multimodal model that runs locally with 16GB of VRAM or unified memory, and Gemma 4 models have crossed 150 million downloads. Google DeepMind |
| WATCHif you work in regulated industries TCS Anthropic partnership. TCS will provide Claude to 50,000 of its own employees across 56 countries and build Claude-powered products for clients in financial services, healthcare, the public sector, and other regulated industries. Anthropic |
| WATCHif you run mission-critical enterprise systems DXC Anthropic alliance. DXC will train tens of thousands of Claude-certified forward-deployed engineers and used Claude to write more than 95% of the code for DXC OASIS, its new AI-native orchestration platform now serving over 50 customers. Anthropic |
| WATCHif you research AI economic impacts Economic Research Exchange. OpenAI launched the OpenAI Economic Research Exchange, a platform for structured project-based collaborations on the economic effects of AI, with applications open until July 5, 2026. OpenAI |
| DEPLOYif you run tutoring or education products Preply Lesson Insights. Preply launched Lesson Insights, an OpenAI API-powered experience that generates tailored feedback across grammar, vocabulary, and pronunciation after each 1:1 lesson, with 75% of English-language learners actively using it and a 4.7/5 satisfaction rating from more than 300k ratings. OpenAI |
| DEPLOYif you scale AI in regulated banking BBVA ChatGPT Enterprise. BBVA now has more than 100,000 employees using ChatGPT Enterprise globally, with 70%+ weekly active usage, around 3 hours saved per employee per week, and up to 80% efficiency gains in selected workflows. OpenAI |
| DEPLOYif you run financial data workflows LSEG OpenAI deployment. LSEG deployed ChatGPT Enterprise and OpenAI APIs across the organization, reducing product release cycles from 3-6 months to 2 weeks and accelerating customer delivery timelines to about 4 weeks from request to production. OpenAI |
Tools and products
| DEPLOYif you run Nitro v3 apps with workflows Workflow SDK on Nitro. Workflow SDK's native Nitro v3 integration is now in beta, with steps running inside the same bundled runtime as the rest of your app and the Nitro dev server serving the workflow web UI at /_workflow. Vercel |
| WATCHif you run agent harnesses in production HarnessAgent in AI SDK 7. AI SDK 7 introduces HarnessAgent, a single API for running established agent harnesses including Claude Code, Codex, and Pi, available on the AI SDK canary release. Vercel |
Breakthroughs
| DEPLOYif you evaluate models during development olmo-eval. Ai2 released olmo-eval, an evaluation workbench that builds on OLMES and supports agentic and multi-turn evaluation, pairwise model comparison, and per-question analysis with standard errors and minimum detectable effects. Allen Institute for AI |
Left out
- Claude Mythos 5: The page text describes Mythos 5 as initially deployed through Project Glasswing in collaboration with the US government, with broader trust
- Gemini 3.5 Live Translate: the publisher's page does not show it available yet
- Claude Fable 5 on AI Gateway: the publisher's page does not show it available yet
- Kimi K2.7 Code: the publisher's page does not show it available yet
- Vercel Drop: the publisher's page does not show it available yet
- DeepSeek via Azure: a figure in the draft could not be found on the publisher's own page
- Threshold billing for Pro teams: the publisher's page does not show it available yet
- GitHub CI on HF Jobs: the publisher's page does not show it available yet
- Nextdoor Codex: the publisher's page does not show it available yet
- OpenAI Academy courses: it did not carry a number that moved
- Domain search CLI: it did not carry a number that moved
- API key budgets: it did not carry a number that moved
- Vercel plugin in Grok Build: it did not carry a number that moved
- OpenAI on Oracle Cloud: it did not carry a number that moved
- Waymo Premier: the condition it attaches to the call did not hold up
The pattern
The week's releases collapse the distance between frontier capability and existing engineering practice. Claude Fable 5 delivers state-of-the-art performance on the longest tasks at $10 per million input tokens and $50 per million output tokens, less than half the price of Claude Mythos Preview, while Gemma 4 12B runs locally with 16GB of VRAM or unified memory and has crossed 150 million downloads. The same model that completed a 50-million-line Ruby codebase migration in a day for Stripe is now being routed into regulated industries through TCS across 56 countries and DXC, which used Claude to write more than 95% of the code for DXC OASIS now serving over 50 customers.
That production pull is matched by tooling that embeds agents inside existing stacks: Workflow SDK's native Nitro v3 integration is in beta with steps in the same bundled runtime, and AI SDK 7 introduces HarnessAgent as a single API for Claude Code, Codex, and Pi. Enterprise deployments show the compounding effect: BBVA has more than 100,000 employees on ChatGPT Enterprise with 70%+ weekly active usage and up to 80% efficiency gains in selected workflows, LSEG cut product release cycles from 3-6 months to 2 weeks, and Preply's Lesson Insights holds a 4.7/5 satisfaction rating from more than 300k ratings. Meanwhile OpenAI's Economic Research Exchange, open until July 5, 2026, and Ai2's olmo-eval workbench turn the model development loop itself into a measured, comparable process.