Overview
- Anthropic has launched an on-demand memory feature for Max, Team and Enterprise subscribers that is turned on by default but can be disabled and does not build a persistent user profile.
- Claude Sonnet 4 now supports up to one million tokens of context in public beta through the Anthropic API and Amazon Bedrock, with Google Cloud’s Vertex AI support coming soon.
- Access to the expanded context window is initially limited to high-tier API customers with Tier 4 or custom rate limits, with broader availability expected in the coming weeks.
- Anthropic raised API rates for prompts exceeding 200,000 tokens to account for higher compute costs, though prompt caching and batch processing can help reduce latency and expenses.
- These privacy-forward and enterprise-focused upgrades aim to boost Claude’s utility for large-scale coding, document synthesis and long-horizon agent workflows in competition with OpenAI and Google.