Claude API changelog

Claude API launches Usage and Cost Analytics API in beta

New analytics endpoints report token usage and cost for Console and Enterprise organizations, grouped by model, workspace, API key or user. The older Usage and Cost API is now legacy but remains supported.

Who is affected:

Admins who track Claude spend and usage through the older reporting endpoints.

What to do:

Plan a move from the legacy Usage and Cost API to the new analytics endpoints.

Claude API raises Messages request size limit from 32 MB to 128 MB

The maximum request size for the Messages API and the Token Counting API rose from 32 MB to 128 MB on the first-party Claude API.

Anthropic offers government partners a 40% discount on Claude Enterprise and API

A new GSA agreement gives government partners a 40% discount on Claude Enterprise, Claude for Government, and the Claude API.

Who is affected:

Government agencies and partners buying Claude through the GSA agreement.

What to do:

Government buyers can check the GSA agreement for eligibility and pricing.

Full Claude Platform available on Google Cloud (Anthropic managed)

Claude Models on Google Cloud (Anthropic managed) is generally available with US and Global endpoints. It gives Google Cloud customers the full Claude Platform feature set with Google Cloud IAM, billing and commitment drawdown. Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 are available, and fast mode for Opus 5.5 costs $8 / $40 per million input and output tokens there.

Who is affected:

Teams that buy Claude through Google Cloud and want features that the Google-managed offering lacks.

What to do:

No action needed. Google Cloud customers who need same-day Claude API features can switch to the Anthropic-managed offering.

API version 2026-10-21 removes temperature, top_p and top_k

Anthropic released API version 2026-10-21 and new major SDK versions that send it by default. Under this version, temperature, top_p and top_k return a 400 error for every model, and the SDKs move from the MIT License to Apache 2.0.

Who is affected:

Any app that upgrades to the new SDK majors (Python 2.0.0, TypeScript 1.0.0 and others).

Action needed:

Remove sampling parameters before upgrading the SDK, or pin the older API version.

Accounts that keep trying to extract Claude's reasoning face stricter usage limits

Anthropic said users who repeatedly try to extract Claude's reasoning now get stricter usage limits, and accounts with persistent distillation attempts are banned. The same post restates that blocked pre-output requests have been billed again since September 24.

Who is affected:

Accounts whose prompts repeatedly try to get Claude to reveal its full reasoning.

What to do:

No action needed for normal use. Do not build workflows that try to extract hidden reasoning.

Max and Team plans now include monthly API credits

Claude Max and Team plans now include monthly credits for the Claude Platform API, claimed by linking a Claude Console organization in billing settings.

Who is affected:

Claude Max and Team subscribers who also build on the API.

What to do:

Link a Claude Console organization in billing settings to claim the credits.

Billing resumes for some pre-output refusals

Refusals that arrive before any output are billed again when stop_details.category is bio, frontier_llm or reasoning_extraction, at the rates of the model that ran the request. Refusals in other categories stay free.

Who is affected:

Apps whose requests are refused in the bio, frontier_llm or reasoning_extraction categories.

What to do:

Check refusal volume by category in your usage data.

Python SDK v1.0 removes sampling parameters and the Text Completions API

Version 1.0 of the Anthropic Python SDK requires Python 3.10 or later and removes the legacy Text Completions API and the temperature, top_p and top_k parameters on Messages methods.

Who is affected:

Python apps that upgrade the anthropic package to 1.0.

What to do:

Follow the v1 migration guide before upgrading the Python SDK.

Claude API rate limits raised and tiers consolidated into three

Sonnet and Haiku rate limits now match Opus at every usage tier, and usage tiers were consolidated into Start, Build and Scale. No organization received lower limits.

Who is affected:

All Claude API organizations.

What to do:

Check your new tier and limits in the Claude Console.

Refusals with no output are no longer billed

On the Claude API, a request that returns stop_reason refusal before Claude generates any output is no longer billed.

Claude Platform on AWS launches

Anthropic launched Claude Platform on AWS, the Claude API on Anthropic-managed infrastructure with AWS billing and IAM authentication.

Anthropic postmortem: three infrastructure bugs degraded Claude responses

Anthropic confirmed that three infrastructure bugs intermittently degraded Claude response quality between August and early September 2025, and said it never reduces model quality because of demand or load.