Amazon Web Services and Anthropic boost enterprise AI capabilities

Amazon Web Services announced a major increase in its Amazon Bedrock AgentCore runtime quotas, expanding default limits up to five times. The new limits allow up to 5,000 active concurrent sessions in the US East (N. Virginia) and US West (Oregon) regions and 2,500 in other supported regions, up from 1,000 and 500 previously. Token‑per‑second interaction capacity was raised from 25 to 200 tokens, and the rate for creating new AI agent sessions was quadrupled to 400 TPM. AWS said the changes are intended to reduce administrative overhead and prevent throttling that can disrupt multi‑agent workflows in large‑scale enterprise deployments.

Anthropic released Claude Sonnet 5, a model positioned as substantially stronger in coding, reasoning and multi‑step agentic tasks, while remaining cost‑efficient for production use. At the same time, the company introduced a self‑hosted Claude Code gateway that runs on Amazon Bedrock and Google Cloud Vertex AI. The gateway lets enterprises keep code, credentials and context inside their own cloud tenancy, supporting bring‑your‑own‑cloud architectures, enterprise SSO and audit logging, thereby creating a security‑focused moat against competing hosted solutions.