You may solely bear in mind the context of as much as 1 million tokens in case your Claude Sonnet 4 has been upgraded and is getting used by way of the API. This may occasionally change sooner or later.
That is 5 instances greater than the earlier restrict. Additionally, Claude means it helps remembering over 75,000 strains of code, and even a whole bunch of paperwork in a single session.
Beforehand, you needed to submit particulars to Claude in a small chunk, however that additionally meant that Claude would neglect the context when he reached the restrict. With a context restrict of as much as 1 million, you possibly can construct higher apps, and Claude remembers the code greater than ever.
It’s value noting that the 1 million context restrict is restricted to Sonnet 4. OPUS4.1 nonetheless has older limitations as it’s an costly mannequin.
Solely the API will get a context restrict of 1 million tokens
The brand new context limits are being deployed by way of synthetic APIs for purchasers with tier 4 and customized price limits, with wider availability rolling out over the approaching weeks.
“Amazon Bedrock additionally has a protracted context and can quickly be coming to Google Cloud’s Vertex AI,” Anthropic mentioned.
“With 1M tokens, you construct an agent that masses the complete codebase with all of your dependencies, analyzes a whole bunch of paperwork directly, and maintains context throughout a whole bunch of instrument calls. Pricing adjusts prompts above 200k tokens, however immediate caching can scale back prices and latency.”
Claude’s cellular and net apps will get a token context restrict of 1 million in some unspecified time in the future sooner or later.
