·3 min read
Claude Mythos 5.2 Launch: Access, Pricing and Upgrade Checks

Anthropic released Claude Mythos 5.2 on November 9, 2026. It shares an underlying model with Claude Fable 5.2, which launched on the same day, and is available only to organizations verified by Anthropic. Both have the same standard API rates and context limits. For teams comparing the two, the first questions are eligibility and data policy. Launch announcement
How does Mythos 5.2 differ from Fable 5.2?
Both models target complex reasoning and longer technical tasks. Anthropic applies stronger safeguards to Fable to support broader product and developer use. Mythos uses an organization verification model. For routine code migrations or long-document analysis, start with a Fable or Opus model you already have access to and use task results to assess whether a change is justified. Launch announcement
API specifications and cost
The Claude API model ID is claude-mythos-5-2. It accepts text and images, returns text, and supports a one-million-token context window and up to 128,000 output tokens. Adaptive thinking stays on and the default effort is high. Anthropic also lists Amazon Bedrock, Google Cloud, and Microsoft Foundry. Availability on each platform depends on your organization’s access. Model specifications
Standard rates are $8 per million input tokens and $40 per million output tokens, both 20% below the $10 and $50 rates for 5.1. A workload of 200,000 input tokens and 25,000 output tokens falls from $3.25 to $2.60 at those rates. This example holds usage constant and excludes caching, tools, and other charges. Migration guide
Three checks before moving from 5.1
1.Confirm retention terms. Anthropic generally requires 30-day data retention. Organizations using zero data retention need explicit authorization. Ask the owner of your data policy to confirm the terms that apply to your workspace. Migration guide
2.Check conversation integrity. A request can return HTTP 400 if it changes an earlier system prompt, tool definition, or message and then resends a thinking block based on the original history. Mythos 5.1 on the Claude API already enforced this check. A move from 5.1 on Bedrock deserves separate regression testing. Migration guide
3.Measure task success and cost separately. Replay a set of completed technical tasks, such as verifying field mappings before and after a data transformation. Record test pass rates, manual correction time, and cost per completed task. Unit prices tell you what a fixed volume costs, while output length and retries determine the actual bill.
Frequently asked questions
Can I select Mythos 5.2 in a regular Claude account?
Anthropic restricts Mythos to verified organizations. Check the access granted to your organization. Launch announcement
How much does fast mode cost?
Anthropic lists $16 per million input tokens and $80 per million output tokens. The reported speedup of up to 2.5 times applies to output generation. Total task time also depends on tool waits and review. Launch announcement