Speed and access split the week between frontier models and platform controls
GPT-5.6 Sol Ultrafast, Gemini 3.7 Flash and ChatGPT Business Premium seats, 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-5.6 Sol Ultrafast
DEPLOYif you run time-sensitive production workflows
WHAT SHIPPEDGPT-5.6 Sol on Ultrafast mode runs up to 14x faster than Standard processing and generates up to 750 output tokens per second, launching first in the OpenAI API and powered by Cerebras.
THE CATCHThe speed tier is a limited preview for a select group of customers, so access is not open to everyone who signs up today.
IF YOU RUNincident response, financial research, customer support, commerce, or live research loops where every second matters
SOURCEOpenAI, Previewing Ultrafast mode: GPT-5.6 Sol at up to 14X the speed
Google DeepMind
Gemini 3.7 Flash
DEPLOYif you build coding or agent workflows
WHAT SHIPPEDGemini 3.7 Flash is available at $0.75 per 1M input tokens and $3.75 per 1M output tokens through the end of the year, half the original 3.6 Flash cost per million tokens.
THE CATCHThe introductory price is temporary, so cost calculations change after the end of the year.
IF YOU RUNsoftware engineering, web development, or knowledge work in finance, law, or biosciences
SOURCEGoogle DeepMind, Introducing Gemini 3.7 Flash
OpenAI
ChatGPT Business Premium seats
DEPLOYif you manage a ChatGPT Business workspace
WHAT SHIPPEDPremium seats are now available on ChatGPT Business at $125 per user per month, or $100 per user per month billed annually, with 5x more usage than Standard seats and no five-hour usage limit.
THE CATCHThe promotion to earn workspace credits for your first premium seats has ended.
IF YOU RUNa team where active teammates hit usage limits on Standard seats
SOURCEOpenAI, Premium seats are coming to ChatGPT Business
The Deploy Log
Models and APIs
| WATCHif you do vulnerability research or exploit validation GPT-5.6-Cyber. GPT-5.6-Cyber completes 95.0% of advanced cybersecurity requests compared with 1.5% for GPT-5.6 Sol, and is available through Daybreak Red access. OpenAI |
Tools and products
| WATCHif you need a 500K context window Grok 4.6 on AI Gateway. Grok 4.6 from SpaceXAI is now available on AI Gateway with a 500K token context window and text and image inputs. Vercel |
| DEPLOYif you already call DeepSeek V4 Pro DeepSeek V4 Pro updated weights. DeepSeek V4 Pro now runs on updated weights on AI Gateway, used by default when you call deepseek/deepseek-v4-pro. Vercel |
| DEPLOYif you use web search in AI Gateway Exa web search free. Exa web search is now free on AI Gateway through August 31, and it is now the default web search for eve agents. Vercel |
| DEPLOYif you manage Vercel Connect from the CLI Vercel CLI 100+ services. You can now integrate 100+ services through Vercel Connect from the CLI, passing the service name to create a connector and attach it to your project. Vercel |
| DEPLOYif you run Vercel Sandbox workloads Vercel Managed Images. Vercel Managed Images introduces versioned, open-source base images, with new sandboxes defaulting to vercel/sandbox/universal:latest on Ubuntu 26.04 with Node.js 24 and Python 3.14. Vercel |
Left out
- Node.js one-click upgrade: the condition it attaches to the call did not hold up
- Bun.serve entrypoint: it did not carry a number that moved
- Vercel Connect observability: it did not carry a number that moved
- LaunchDarkly on Vercel Marketplace: it did not carry a number that moved
- Free domain for Pro teams: the condition it attaches to the call did not hold up
- AI Gateway coding agents setup: it did not carry a number that moved
- ACP harness support: it did not carry a number that moved
- Grok Build harness adapter: it did not carry a number that moved
- GLM 5.2 free for eve agents: its own name for the thing was too long to print
- Gemini 3.7 Flash on AI Gateway: its own name for the thing was too long to print
- Enterprise Managed Users: it did not carry a number that moved
- Certificate Transparency Monitoring: its own name for the thing was too long to print
- Cloudflare Access on Workers: it did not carry a number that moved
- Vercel CDN ECH: it did not carry a number that moved
- SL2T sign language model: the condition it attaches to the call did not hold up
- Daybreak on AWS: it did not carry a number that moved
The pattern
The week splits between frontier speed and platform plumbing. OpenAI ships GPT-5.6 Sol Ultrafast at up to 750 output tokens per second and 14x Standard speed, but only as a limited preview for a select group of customers, while Google DeepMind prices Gemini 3.7 Flash at $0.75 per 1M input tokens and $3.75 per 1M output tokens through the end of the year, half the original 3.6 Flash cost per million tokens. ChatGPT Business Premium seats open at $125 per user per month, or $100 per user per month billed annually, with 5x more usage than Standard seats and no five-hour usage limit, yet the workspace credits promotion has ended. GPT-5.6-Cyber completes 95.0% of advanced cybersecurity requests compared with 1.5% for GPT-5.6 Sol, available only through Daybreak Red access.
Vercel turns the same week into a control surface for everything else. Grok 4.6 from SpaceXAI lands on AI Gateway with a 500K token context window and text and image inputs, DeepSeek V4 Pro runs on updated weights by default when you call deepseek/deepseek-v4-pro, and Exa web search is free on AI Gateway through August 31 as the default web search for eve agents. Vercel CLI now integrates 100+ services through Vercel Connect, and Vercel Managed Images introduces versioned, open-source base images, with new sandboxes defaulting to vercel/sandbox/universal:latest on Ubuntu 26.04 with Node.js 24 and Python 3.14. The frontier models gate speed and capability behind previews, temporary prices, or Daybreak Red access, while the platform layer makes routing, search, and base images default and free.