GPT-6 Intelligent UI puts interactive answers before 1.2 billion weekly ChatGPT users
GPT-6 Intelligent UI, Claude Haiku 5.5 and Clef-omni, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
OpenAI
GPT-6 Intelligent UI
DEPLOYif you build ChatGPT workflows
WHAT SHIPPEDGPT-6 with Intelligent UI rolls out globally to ChatGPT Plus, Pro, Business and Enterprise tiers today, expanding to Free and Go tiers tomorrow. The model composes responses with graphics, buttons, forms, charts and interactive experiences, and GPT-6 Instant starts answering web search questions 44% sooner than GPT-5.6 Instant.
THE CATCHThe models powering Work and Codex are not changing as part of this release, so agentic coding workflows see no model change. Enterprise availability depends on workplace admin settings.
IF YOU RUNa support or education product on ChatGPT, you can now return tappable tools and interactive diagrams instead of text
SOURCEOpenAI, GPT-6 and Intelligent UI for everyone
Amazon Web Services
Claude Haiku 5.5
DEPLOYif you run high-volume subagents
WHAT SHIPPEDClaude Haiku 5.5 is available today on Amazon Bedrock through US, EU, AU, JP and Global inference profiles, and costs around 75 percent less than Claude Haiku 4.5 for most tasks. It is the first Haiku model with effort controls for tuning cost against intelligence per task.
THE CATCHHaiku 5.5 is positioned as the fast layer under Opus 5.5, which plans and makes judgment calls while Haiku handles routing, classification, summarization and small code changes across many files.
IF YOU RUNparallel review or routing subagents, you can cut cost per task and tune effort per workload
SOURCEAmazon Web Services, Introducing Claude Haiku 5.5 on AWS
Cloudflare
Clef-omni
DEPLOYif you classify audio or video
WHAT SHIPPEDCloudflare released Clef-omni, an open-weight decision model that takes audio in wav or mp3, video in mp4 or webm, image and text input. Cloudflare also cut the price of Clef-flash so it is now cheaper than Jev, and made Clef faster.
THE CATCHClef-omni removes the need for cascading pipelines that transcribe speech to text or split audio and image channels from video, letting one model make decisions across modalities.
IF YOU RUNmoderation or inspection pipelines on media, you can replace multi-model transcription and splitting chains with one call
The Deploy Log
Models and APIs
| DEPLOYif you verify media provenance SynthID Detector. Google expanded SynthID Detector globally in English, letting anyone check if an image, video or audio was made with AI from Google or partners including OpenAI, NVIDIA, Kakao and soon Apple, after watermarking over 180 billion images and videos and 240,000 years of audio. Google |
| WATCHif you code on Windows MAI Code 1.1 Flash. Microsoft AI developed a local version of MAI Code 1.1 Flash, a 137 billion total and 6.8 billion active parameter mixture-of-experts model, with the quantized on-device version coming in at 53GB, an 80% reduction in size, and scoring 70.80% on SWE-Bench Verified. Microsoft |
| DEPLOYif you run a security operations center Sophos Daybreak. Sophos cut average response time for cases using agents from approximately 38 minutes to 89 seconds, a 96% reduction, and now resolves 52% of MDR cases end-to-end with AI built through OpenAI Daybreak. OpenAI |
| DEPLOYif you run browser automation Asana browser agent. Asana used GPT-6 Astra in Codex to optimize StackAI browser agent on GPT-6.1 Sol, cutting estimated model cost 76x to $0.47 per run and making runs 5x faster at roughly four minutes, with 89% of input from cache at 5% of the uncached price. OpenAI |
Tools and products
| DEPLOYif you classify or score requests Microsoft Decision-1. Microsoft Decision-1 is now available on AI Gateway, answering typed questions about text with probabilities, choices and scores for classification, workflow routing or rubric assessment, with no separate Azure account or deployment required. Vercel |
| DEPLOYif you need visual classification Liquid d1. Liquid d1 is now available on AI Gateway, a decision model that evaluates shared state against typed questions for classification, routing and scoring, returns structured answers with probabilities without generating text tokens, and supports vision for image classification and inspection. Vercel |
| DEPLOYif you generate short video clips Grok Imagine Video 1.5 Lite. Grok Imagine Video 1.5 Lite from xAI is now available on AI Gateway, turning text prompts and images into video with native audio in clips from 1 to 15 seconds at 480p, 720p and 1080p across seven aspect ratios. Vercel |
| DEPLOYif you need faster interactive output OpenAI Ultrafast tier. AI Gateway now supports OpenAI's Ultrafast service tier for GPT 6 Astra and GPT 6.1 Sol, billed at 6x the standard per-token rate, with requests pinned to unsupported regions such as the EU running at the standard tier. Vercel |
| DEPLOYif you generate or edit images FLUX 3 Image. FLUX 3 Image from Black Forest Labs is now available on AI Gateway, generating images from text or editing existing images with up to ten reference images and five resolution tiers from 768 by 768 to 4K, with no Black Forest Labs account required. Vercel |
| WATCHif you work with large codebases Step 5 Preview. Step 5 Preview from StepFun is now available on AI Gateway, built for agentic coding, professional knowledge work and financial analysis, with text and image input and a 1M-token context window. Vercel |
| DEPLOYif you use Copilot across accounts GitHub Copilot weekly releases. GitHub Copilot added Claude Haiku 5.5 for Pro, Pro+, Max, Business and Enterprise users, made local sandboxing generally available in Copilot CLI, the Copilot app and VS Code sessions using Agent Host, and let the Copilot app use separate GitHub accounts for license and repositories. GitHub |
| DEPLOYif you run GitHub code scanning CodeQL 2.27.2. CodeQL 2.27.2 is now available, adding a C++ regular-expression parser, SQL-injection sink models for the Comdb2 C API, Go websocket modeling, Rust async data flow improvements, and JavaScript Workflow SDK directive recognition, with the Default suite running 498 security queries covering 170 CWEs. GitHub |
| DEPLOYif you run Claude Code in production Claude Code v2.1.296. Claude Code v2.1.296 added a code key to the Claude apps gateway managed policies, an autoCompactWindow option for subagents, a CLAUDE_CODE_WORKFLOW_SUBAGENT_MODEL environment variable, and updated Sonnet 5.5 cache reads to $0.10 per million tokens from $0.20. Anthropic |
| DEPLOYif you use Codex TUI Codex rust-v0.162.1. Codex rust-v0.162.1 fixed a TUI crash when asynchronous questions contain multiple lines, preserving line breaks and complete hyperlink destinations, and fixed startup failures from differences between a running background server's feature settings and CLI defaults. OpenAI |
| DEPLOYif you run local models Ollama v0.40.2. Ollama v0.40.2 upgrades models downloaded with earlier versions in the background the first time you run them for better performance and compatibility on llama.cpp, keeps the original copy as a backup, and makes ollama list stop showing duplicate entries for upgraded models. Ollama |
| DEPLOYif you run Google Voice in Europe Google Voice Carrier Link. Google made Carrier Link available in 14 additional countries including the US, Canada and much of Europe, letting Workspace customers add phone numbers and calling plans from certified local carriers through partners Southlight Services, Gamma Communications, Tata Communications and Telefonica Global Solutions. Google Workspace |
| DEPLOYif you build Workspace automations Workspace Studio October release. Google Workspace Studio added nested conditionals, Gmail skip replies for thread triggers, Google Meet calendar outputs with full invitee lists and post-meeting email draft templates, and a Search for a Doc Drive step, rolling out gradually starting October 13, 2026. Google Workspace |
Left out
- Vercel CLI domain purchases: it did not carry a number that moved
- Workers on-demand profiling: it did not carry a number that moved
The pattern
The week's releases split into two motions: OpenAI and AWS pushed frontier and fast models to the largest user bases, while Cloudflare, Vercel and GitHub shipped the connective tissue that makes those models usable in production. GPT-6 Intelligent UI reaches 1.2 billion weekly ChatGPT users, Claude Haiku 5.5 lands on Bedrock at 75% lower cost than its predecessor, and Clef-omni adds audio and video to decision models, all in the same window.
The second motion is about control and cost. Sophos and Asana show the same pattern from opposite directions: Sophos cut investigation time from 38 minutes to 89 seconds with Daybreak agents, while Asana cut browser agent cost 76x to $0.47 per run by fixing caching and history management. Vercel's decision models, Cloudflare's on-demand profiling, and GitHub's sandboxing all give operators the levers to run those agents safely and cheaply.