Anthropic will remove fast mode for Claude Opus 4.8 on November 30, 2026
On October 30, 2026, Anthropic marked fast mode for Claude Opus 4.8 as deprecated. Once it is removed, fast requests to that model will return an error instead of running at standard speed. The company points users to Claude Opus 5.5, where fast mode costs $8 per million input tokens and $40 per million output tokens.
Anthropic deprecated fast mode for Claude Opus 4.8 on October 30, 2026, and will remove it on November 30, 2026. Fast mode is a paid setting that runs the same model at a higher output speed. After removal, requests for fast mode on Opus 4.8 will fail, so developers and Claude Code users who rely on it need to move to Claude Opus 5.5 to keep the speedup.
What stops working
On the Claude API, fast mode is turned on with speed: "fast" in the request. After November 30, 2026, a request to claude-opus-4-8 with that setting will return an error. Anthropic's release notes for Claude Code 2.1.316, published on October 31, 2026, add that such requests will not fall back to standard Opus 4.8 either.
Fast mode on Opus 4.8 was a research preview, a status Anthropic uses for features whose availability and pricing may still change. Claude Console organizations needed provisioned access before they could use fast mode on that model.
Anthropic has followed the same schedule before. Fast mode for Opus 4.7 was deprecated on June 25, 2026, and removed on July 24, 2026.
What the switch costs
Fast mode is billed at a premium on every model that supports it. On Opus 4.8 it costs $10 per million input tokens and $50 per million output tokens, twice the standard price of $5 and $25. On Opus 5.5 it costs $8 and $40, also twice that model's standard price of $4 and $20. For a team already paying for fast mode, moving to Opus 5.5 therefore lowers the price per token.
Anthropic says fast mode delivers up to 2.5 times more output tokens per second than standard speed and does not change the model's capabilities. Opus 5.5 also has a default fast mode rate limit in the Start, Build, and Scale usage tiers, so organizations in those tiers do not need to request access.
In Claude Code, fast mode is toggled with the /fast command. Subscribers on the Pro, Max, Ultra, Team, and Enterprise plans pay for it only through usage credits, outside the limits included in their plan.
Sources
Once a week, in your inbox
The whole week's news, gathered into one email. No ads, no passing your address on, one-click unsubscribe.
The address is used only for this roundup. It is not sold and not passed to anyone.
Related
ChatGPT Pro and Business plans now include up to $500 a month in OpenAI API credits
OpenAI now gives eligible ChatGPT Pro and Business subscribers monthly credits for its API. A Pro plan returns its own price in credits, and a Business workspace gets up to $500 a month. The credits expire at the end of each billing cycle.
OpenAIverified
Mistral releases the weights of Mistral Large 4, but selling hosted access needs a separate license
Mistral has made its largest model downloadable. The license is a modified MIT license that allows internal and commercial use, while a business that offers the model to others as a service has to sign a separate agreement. Mistral says the released version codes better and handles defensive security work better than the preview.
Mistral AIverified
NVIDIA releases Nemotron 3.5 Transcribe, an open model that transcribes 40 languages and labels speakers
NVIDIA published Nemotron 3.5 Transcribe with open weights and a license that allows commercial use. One model transcribes 40 languages, translates 31 of them to or from English, and tells speakers apart in 13. The full-precision checkpoint needs one 80 GB H100, and the smaller quantized versions are announced but not yet out.
NVIDIAverified