Claude Sonnet 4 now supports 1M tokens of context (www.anthropic.com)

Anthropic has announced that Claude Sonnet 4 now supports an unprecedented 1 million tokens of context—quintupling its previous limit—available in public beta via the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI. This massive context window enables processing of entire large-scale codebases with over 75,000 lines of code or dozens of lengthy documents in a single request, unlocking new possibilities for AI-driven code analysis, document synthesis, and multi-step context-aware agent workflows. For developers, this means Claude can deeply understand complex project architectures, identify cross-file dependencies, and provide holistic suggestions across entire systems, while also maintaining coherence in lengthy multi-tool interactions. The expanded context also empowers large-scale document analysis, such as parsing extensive legal contracts or scientific literature, enabling insights that span vast information sets. Pricing adjusts for usage beyond 200K tokens to cover increased computation costs, but features like prompt caching and batch processing can optimize both cost and latency. Early adopters like Bolt.new and iGent AI highlight the transformative impact on real-world coding and autonomous software engineering agents, enabling multi-day development sessions and scalable production workflows that were previously impractical. With the rollout across major cloud platforms and exploration of broader product integration underway, Claude Sonnet 4’s one-million-token context sets a new milestone for large-context AI applications in software development and beyond.