changelogAnthropic

Anthropic Python SDK v2.7.0 adds claude-fable-5-2 and claude-mythos-5-2 model IDs

TL;DR

Version 2.7.0 of the Anthropic Python SDK adds API support for claude-fable-5-2 and claude-mythos-5-2. It also adds parent organization filtering to usage and cost reports, a new webhook event, and a set of Bedrock, Vertex and Foundry client fixes.

2 min read
0

Anthropic's Python SDK v2.7.0 adds API support for two model identifiers, claude-fable-5-2 and claude-mythos-5-2. The release also extends the Admin API and usage reporting, and fixes a series of bugs in the Bedrock, Vertex and Foundry clients.

Features

  • New model IDs: claude-fable-5-2 and claude-mythos-5-2 are added to the API.
  • Usage and cost reports: parent organization filtering and grouping are added to both report types.
  • Webhooks: organization webhook endpoints now accept user_profile.program_status_updated, and the user_profile.program_status_updated webhook event is added.
  • Admin API: the sandbox_egress_restricted error is added to Admin API write methods.

Bug fixes

Bedrock and Vertex

  • The parse and count_tokens methods, previously missing from beta.messages.with_raw_response and with_streaming_response, are added.
  • The X-Stainless-Async header is now sent, and raw requests that name no stream class are now streamed.

Bedrock

  • Requests with no Connection header are now signed instead of raising KeyError.
  • On the Bedrock Mantle clients, a list in a query is now written as a[]=1&a[]=2, matching Anthropic's behavior.

Client

  • API errors can now be unpickled and copied.
  • ANTHROPIC_CUSTOM_HEADERS is now read on the Bedrock, Bedrock Mantle and Vertex clients.

Foundry

  • copy, with_options and with_middleware no longer raise when ANTHROPIC_FOUNDRY_RESOURCE is set.

Parsing

  • const is now kept in structured output schemas, and type is dropped beside const and enum.

Chores

  • Some error detail types are removed and descriptions are updated.
  • Documentation now says when a work heartbeat does not compare its value.
  • The repository's review rules are added, and two comments are removed from the error classes.

What this means

Developers who call the API through the Python SDK can now reference claude-fable-5-2 and claude-mythos-5-2 by those identifiers. Teams running Claude through Bedrock, Bedrock Mantle, Vertex or Foundry get the most fixes in this release: request signing without a Connection header, custom header support through ANTHROPIC_CUSTOM_HEADERS, and Foundry client methods that no longer fail when ANTHROPIC_FOUNDRY_RESOURCE is set. Organizations using the Admin API gain parent organization filtering and grouping in usage and cost reports, along with a new webhook event for user_profile.program_status_updated. The release listing gives no pricing, context window or benchmark data for the two new model IDs.

Related Articles

changelog

Claude Fable 5.2 is now generally available in GitHub Copilot

Claude Fable 5.2, described as Anthropic's newest Fable model, is now generally available in GitHub Copilot. GitHub says it can be used for complex, multistep coding tasks across large codebases.

product update

Claude Fable 5.2 and Mythos 5.2 now available on Amazon Bedrock and Claude Platform on AWS

AWS announced that Claude Fable 5.2 and Claude Mythos 5.2 are available on Amazon Bedrock and Claude Platform on AWS. Anthropic describes Fable 5.2 as its most capable generally available model for technical work, while Mythos 5.2 is limited to Anthropic's verification programs for cyber security and life sciences.

model release

Anthropic releases Claude Fable 5.2 and Mythos 5.2 at $8/$40 per 1M tokens, with an optional fast mode

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, priced at $8 per million input tokens and $40 per million output tokens. Anthropic says they generate output up to 30% faster than the 5.1 models, and Fable 5.2 reaches 49.1% on FrontierCode 2.0 at Xhigh effort.

analysis

Seth, Block and Tononi argue in Nature that today's AI is not conscious and probably cannot be on silicon

Consciousness researchers Anil Seth, Ned Block and Giulio Tononi argue in a Nature commentary that running the right computations is not enough for consciousness. They respond to Richard Dawkins and Geoffrey Hinton, who consider AI possibly already conscious, and to the Butlin report's 14-indicator checklist.

Comments

Loading...