Frontier models ship everywhere as deployment splits into agents, images, infrastructure
Waymo five new cities, GPT-5.1-Codex-Max and Gemini 3, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
Waymo
Waymo five new cities
DEPLOYif you ride or drive in these metros
WHAT SHIPPEDWaymo is introducing fully autonomous driving in five new cities: Miami, Dallas, Houston, San Antonio, and Orlando. Operations start today in Miami, and will begin in the remaining four cities over the coming weeks, ahead of opening doors to riders next year.
THE CATCHThe expansion is framed as routine, not a technical feat. Waymo says the Waymo Driver is involved in 11 times fewer serious injury collisions compared to human drivers, and that local nuances are becoming fewer with every city.
IF YOU RUNa fleet or transit service in Texas or Florida and need a reference point for rider-only operations
SOURCEWaymo, Safe, Routine, Ready: Autonomous driving in five new cities
OpenAI
GPT-5.1-Codex-Max
DEPLOYif you run long coding tasks
WHAT SHIPPEDGPT-5.1-Codex-Max is available in Codex today for the CLI, IDE extension, cloud, and code review. It is the first model natively trained to operate across multiple context windows through compaction, working over millions of tokens in a single task.
THE CATCHOn SWE-bench Verified, GPT-5.1-Codex-Max with xhigh reasoning effort reaches 77.9%, up from 73.7% for GPT-5.1-Codex high. OpenAI says 95% of OpenAI engineers use Codex weekly and ship roughly 70% more pull requests since adopting Codex.
IF YOU RUNa team doing project-scale refactors or multi-hour agent loops and you want a model that compacts its own context
SOURCEOpenAI, Building more with GPT-5.1-Codex-Max
Google DeepMind
Gemini 3
DEPLOYif you build with Google models
WHAT SHIPPEDGemini 3 Pro is available today in the Gemini app, AI Studio, Vertex AI, and Google Antigravity. It tops the LMArena Leaderboard with a score of 1501 Elo and scores 76.2% on SWE-bench Verified.
THE CATCHGemini 3 Pro scores 37.5% on Humanity's Last Exam without tools and 91.9% on GPQA Diamond. Gemini 3 Deep Think reaches 41.0% on Humanity's Last Exam and 45.1% on ARC-AGI-2 with code execution, but is only open to safety testers before Google AI Ultra subscribers.
IF YOU RUNa product team choosing between Gemini 3 Pro and the previous generation for reasoning-heavy work
SOURCEGoogle DeepMind, A new era of intelligence with Gemini 3
The Deploy Log
Models and APIs
| DEPLOYif you make ads or mockups Nano Banana Pro. Nano Banana Pro, built on Gemini 3 Pro, generates images with legible text in multiple languages, blends up to 14 images, maintains consistency of up to 5 people, and supports 2K and 4K resolution. Google DeepMind |
| DEPLOYif you are on Azure Claude in Microsoft Foundry. Claude Sonnet 4.5, Haiku 4.5, and Opus 4.1 are now available in public preview in Microsoft Foundry, and Microsoft's Agent Mode in Excel includes an option to use Claude in preview. Anthropic |
| WATCHif you use Intuit products Intuit OpenAI partnership. OpenAI and Intuit entered a multi-year strategic partnership with a new $100M+ agreement to deepen Intuit's use of OpenAI frontier models across TurboTax, Credit Karma, QuickBooks, and Mailchimp. OpenAI |
| DEPLOYif you want agent-first development Google Antigravity. Google Antigravity is available in public preview at no charge with generous rate limits on Gemini 3 Pro usage, and includes access to Gemini 3, Claude Sonnet 4.5, and OpenAI's GPT-OSS. Google DeepMind |
| DEPLOYif you fine-tune with TRL RapidFire AI. Hugging Face TRL now officially integrates with RapidFire AI to accelerate fine-tuning and post-training experiments, with internal benchmarks showing approximately 16-24x higher experimentation throughput than sequential config comparison. Hugging Face |
Tools and products
| DEPLOYif you route through Vercel Gemini 3 Pro. Gemini 3 Pro is now accessible via Vercel's AI Gateway with no other provider accounts required, using the model ID google/gemini-3.0-pro-preview. Vercel |
| DEPLOYif you need agentic tool calling Grok 4.1 Fast. Grok 4.1 Fast Reasoning and Grok 4.1 Fast Non-Reasoning are now accessible via Vercel's AI Gateway with no other provider accounts required, and both have a 2M context window. Vercel |
| DEPLOYif you use Notion Agent Gemini 3 Pro in Notion. Gemini 3 Pro is now available in Notion, giving the Notion Agent access to Google's latest model alongside those from OpenAI and Anthropic. Notion |
| DEPLOYif you run Pro or Enterprise Vercel Agent investigations. Vercel Agent investigations are now included in Observability Plus, adding 10 investigations to every billing cycle at no extra cost to the subscription. Vercel |
| DEPLOYif you monitor app security Vercel Firewall UI. Vercel launched improvements to the Firewall UI, including an updated Overview page for DDoS attacks and a new Traffic page to drill into top sources of traffic by IPs, request paths, JA4 digests, ASN, and user agents. Vercel |
Breakthroughs
| DEPLOYif you run multilingual meetings Real-time speech translation. Google Research introduced an end-to-end speech-to-speech translation model that enables real-time translation in the original speaker's voice with only a 2-second delay, now available in Google Meet on servers and as an on-device feature for Pixel 10 devices. Google Research |
| DEPLOYif you route EV drivers EV charging port prediction. Google Research deployed a lightweight linear regression model that predicts EV charging port availability and reduces bad predictions by approximately 20% in morning peak times and approximately 40% in evening peak times. Google Research |
Left out
- WeatherNext 2: the publisher's page does not show it available yet
- Open ASR Leaderboard: the publisher's page does not show it available yet
- Generative UI: it did not carry a number that moved
- AnyLanguageModel: it did not carry a number that moved
- CDN cache invalidation: it did not carry a number that moved
- Elysia on Vercel: it did not carry a number that moved
- FastAPI and Flask build commands: it did not carry a number that moved
- Nano Banana Pro: it repeats an item this edition already carries
- Target app in ChatGPT: it did not carry a number that moved
- ChatGPT for Teachers: the condition it attaches to the call did not hold up
The pattern
Frontier models no longer ship as isolated releases. They ship as distribution events. Gemini 3 Pro appears in the Gemini app, AI Studio, Vertex AI, Google Antigravity, Vercel's AI Gateway, and Notion on the same day it tops LMArena with 1501 Elo. Claude Sonnet 4.5, Haiku 4.5, and Opus 4.1 enter Microsoft Foundry and Microsoft 365 Copilot. GPT-5.1-Codex-Max lands in Codex for CLI, IDE extension, cloud, and code review. The model is not the product; the product is the surface where a team already works.
That surface is splitting into three layers. Agents handle project-scale refactors and multi-hour loops, with GPT-5.1-Codex-Max compacting its own context and OpenAI engineers shipping roughly 70% more pull requests. Images become a reasoning output, as Nano Banana Pro blends up to 14 images and holds consistency for up to 5 people at 2K and 4K. Infrastructure absorbs the rest: Vercel Agent investigations add 10 investigations per billing cycle in Observability Plus, the Vercel Firewall UI tracks DDoS attacks and JA4 digests, and Google Antigravity gives Gemini 3 Pro away at no charge. Even physical deployment follows the same logic. Waymo opens five new cities in Miami, Dallas, Houston, San Antonio, and Orlando, treating the expansion as routine rather than a technical feat. The $100M+ Intuit agreement and the 2-second speech-to-speech delay in Google Meet are not separate stories. They are the same story: the frontier is now a logistics problem.