Claude Sonnet 4 has been upgraded, and it may well now keep in mind as much as 1 million tokens of context, however solely when it is used by way of API. This might change sooner or later.
That is 5x greater than the earlier restrict. It additionally signifies that Claude now helps remembering over 75,000 traces of code, and even a whole lot of paperwork in a single session.
Beforehand, you have been required to submit particulars to Claude in small chunks, however that additionally meant Claude would neglect the context because it hit the restrict. With as much as a 1 million context restrict, you’ll be able to construct higher apps, and Claude can keep in mind extra of your code than ever.
It’s value noting that the 1 million context restrict is proscribed to Sonnet 4. Opus 4.1 nonetheless has the outdated limitations as a result of it is an costly mannequin.
Solely API will get 1 million tokens context restrict
The brand new context restrict is rolling out by way of the Anthropic API for patrons with Tier 4 and customized price limits, with broader availability rolling out over the approaching weeks.
“Lengthy context can also be accessible in Amazon Bedrock and is coming quickly to Google Cloud’s Vertex AI,” Anthropic famous.
“With 1M tokens you’ll be able to: load complete codebases with all dependencies, analyze a whole lot of paperwork directly, and construct brokers that preserve context throughout a whole lot of software calls. Pricing adjusts for prompts over 200K tokens, however immediate caching can scale back prices and latency.”
Claude’s cellular and net apps shall be getting the 1 million token context restrict sooner or later sooner or later.
46% of environments had passwords cracked, almost doubling from 25% final yr.
Get the Picus Blue Report 2025 now for a complete take a look at extra findings on prevention, detection, and information exfiltration developments.
