Anthropic AI Updates: October 31, 2026

1. Anthropic Deprecates Fast Mode on Claude Opus 4.8, Removal Set for November 30

Anthropic deprecated fast mode for Claude Opus 4.8 on October 30, 2026 and will remove it on November 30, 2026. After removal, Messages API requests to claude-opus-4-8 with speed: "fast" return an error instead of falling back to standard speed, so callers that rely on the SDK’s fallback pattern should check their error handling before the cutoff. Fast mode runs the same model with a faster inference configuration for up to 2.5x higher output tokens per second at a price premium: $10 per million input tokens and $50 per million output tokens on Opus 4.8 and Opus 5, versus $8 and $40 on Opus 5.5. The recommended migration is Claude Opus 5.5, where fast mode is generally available without a beta header and has default rate limits in the Start, Build, and Scale usage tiers; Opus 5 keeps fast mode in research preview behind the fast-mode-2026-02-01 header. Source

2. Usage and Cost Analytics API Launches in Beta for Console and Enterprise Organizations

Anthropic released the Usage and Cost Analytics API in beta on October 29, gated by the anthropic-beta: analytics-usage-cost-2026-08-24 header. GET /v1/organizations/analytics/usage returns token counts by type (uncached input, cache creation, cache read, output) plus request counts, and GET /v1/organizations/analytics/cost returns USD amounts, with rows grouped by up to five dimensions such as model, workspace_id, api_key_id, service_tier, context_window, inference_geo, speed, and actor, the user, service account, or federated AWS or Google Cloud identity behind each request. Results can be split into 1m, 1h, or 1d UTC buckets or ranked by tokens, requests, or cost, and each request covers at most 31 days starting no earlier than January 1, 2026, with data typically refreshed within 4 hours and subject to revision until about 7 days after month end. The same endpoints serve Claude Console organizations (Admin API key or org:admin OAuth token) and Claude Enterprise organizations (a read:analytics key), and they are also exposed on Claude Platform on AWS and Claude Models on Google Cloud with native IAM authorization. The existing Usage and Cost API and the Enterprise usage_report, cost_report, user_usage_report, and user_cost_report endpoints are now legacy but remain supported, and Anthropic published a field-by-field migration guide. Source

3. Claude Code 2.1.316 Adds Subagent Effort Control, Layered Gateway Policies, and Bedrock Guardrail Input Tagging

Anthropic shipped Claude Code v2.1.316 with three new environment variables: CLAUDE_CODE_SUBAGENT_EFFORT sets the effort level subagents run at when nothing else names one, instead of inheriting the session’s; CLAUDE_CODE_DISABLE_ULTRATHINK turns the ultrathink keyword off; and CLAUDE_CODE_DISABLE_LOOPED_CROSS_CERT_DROP keeps system cross-certificates that certify each other in a loop, for private CAs that need one. The Claude apps gateway gains pooled group budgets, a shared monthly budget per group on top of each member’s own limit set with pooled_budget on the spend limits admin API, a managed.policy_mode: layered option under which every policy matching a developer applies and is merged over the ones below it rather than only the first match, and guardrail.input_tagging on Bedrock upstreams, which wraps prompts in the input tags that an Amazon Bedrock guardrail’s prompt attacks filter requires (with all_messages_and_system also tagging the system prompt). The VS Code extension can now sign in to a Claude apps gateway from the chat panel without a terminal, cloud sessions show a form card so users can fill in or decline a form an MCP server requests instead of the request being silently canceled, and the Code Review analytics page adds a findings-by-severity card with a daily stacked chart. Source