# GPT-6.1 Sol Ultrafast: now for every API user, also in the EU

https://aycode.ai/blog/gpt-6-1-sol-ultrafast/ · 2026-10-10 · News

Since 8 October every OpenAI API user can run GPT-6.1 Sol in Ultrafast mode, with EU data residency. It costs six times Standard. Codex has it too.

OpenAI's fastest service tier just got a lot more reachable. Since 8 October, according to the API changelog, every API user can run GPT-6.1 Sol in Ultrafast mode, and for the first time with data residency in Europe as well as the US.

## What Ultrafast is

Ultrafast is a service tier, not a new model: you send gpt-6.1-sol with service_tier: "ultrafast" in the Responses API and get less time between output tokens. The guide recommends a persistent WebSocket connection, especially for agents that make many tool calls in quick succession, because otherwise network overhead eats part of the gain. Ultrafast has its own rate limits, separate from Standard and Fast, so check your organization's limits before you move traffic.

The detail that matters in Europe: GPT-6.1 Sol Ultrafast runs with global processing or with US or EU data residency. GPT-6 Astra also has an Ultrafast mode, but only with US residency or global processing.

## What it costs

The pricing page lists GPT-6.1 Sol for prompts up to 272K tokens like this, per million tokens:

- Standard: $2 input, $0.10 cached input, $10 output.
- Fast: $4 input, $0.20 cached input, $20 output.
- Ultrafast: $12 input, $0.60 cached input, $60 output.

So Ultrafast costs six times Standard, and the data residency endpoints add 10% on top. Longer prompts cost more in every tier.

## In Codex and ChatGPT Work

The Codex changelog dates the same speed-up for Codex and ChatGPT Work to 8 October as well. There it is part of the Pro $500 plan and eligible Enterprise and Edu plans. In Enterprise workspaces it is off by default until an owner turns it on, and inference residency is available in the US and in Europe (EEA and Switzerland). GPT-6.1 Sol is already the default for complex work in Codex (our post from 6 October).

## Also new in the API this week

- Decisions API (beta, 6 October): typed answers instead of text, more in our post on decision models.
- Three usage tiers instead of five (6 October): Build, Launch and Grow. Organizations move up automatically as their credit purchases reach each tier's minimum.
- HIPAA support (5 October): admins of eligible organizations can accept the Business Associate Agreement in the API settings.

## Our take

Speed is not free, and six times the price is a real number. Ultrafast can pay off where waiting costs more than tokens: an agent loop with dozens of quick tool calls, or a live session where someone watches every token. Batch work or overnight jobs may benefit less. The EU residency option matters for teams with regional requirements, but they still need to configure the appropriate endpoint and check that the service meets those requirements.

## In AyCode

In a Codex chat in AyCode, the speed menu next to the model offers Normal, Fast and Ultrafast for GPT-6.1 Sol and Astra, and shows how much of your Codex allowance each speed draws: 1×, 2× and 8×. Whether Ultrafast works for you is decided by your Codex plan, as described above. How the model and speed pickers work is in the docs, and how AyCode compares with the Codex app is on AyCode vs Codex.

Sources:
- OpenAI: API changelog (https://developers.openai.com/api/docs/changelog)
- OpenAI: Ultrafast mode (https://developers.openai.com/api/docs/guides/ultrafast-mode)
- OpenAI: API pricing (https://developers.openai.com/api/docs/pricing)
- OpenAI: GPT-6.1 Sol model (https://developers.openai.com/api/docs/models/gpt-6.1-sol)
- OpenAI: Codex changelog (https://learn.chatgpt.com/docs/changelog)