Mythos 5.1 is not a better Fable 5.1. It is the same capability behind a different door.
Anthropic's docs list Mythos 5.1 with the same capabilities, specs and price as Fable 5.1. What actually differs is who gets access and one API check.
GTM Architect & Growth Operator · Now · 16 September 2026
TL;DR · Key insights
- Anthropic's own docs describe Mythos 5.1 as having the same capabilities as Fable 5.1. They share a 1M token context window, 128k max output, always-on adaptive thinking, tokenizer and price.
- The difference that matters is access. Mythos 5.1 is offered only to approved participants in Project Glasswing, Anthropic's programme for securing critical software. Everyone else uses Fable 5.1.
- There is one API difference. Fable 5.1 checks whether you edited conversation history before a thinking block and can reject the request. Mythos 5.1 does not run that check.
- If you are not in Glasswing, the comparison has a one-word answer. If you are, test your integration on Fable 5.1 anyway, because the history check is the thing that breaks on the public model.
People keep asking which of the two is better. Anthropic’s own documentation answers it in one line, in the model table where most readers only look at the ID column.
The description for Claude Mythos 5.1 reads: same capabilities as Claude Fable 5.1.
| Info | Claude Fable 5.1 | Claude Mythos 5.1 |
|---|---|---|
| Capabilities | Baseline | Same as Fable 5.1 |
| Input / output per 1M | $10 / $50 | $10 / $50 |
| Cache read per 1M | $0.25 | $0.25 |
| Context / max output | 1M / 128K | 1M / 128K |
| Thinking | Adaptive, always on | Adaptive, always on |
| Forced tool use | Not supported | Not supported |
| History-editing check | Runs | Does not run |
| Who can use it | All customers | Project Glasswing only |
| API ID | claude-fable-5-1 | claude-mythos-5-1 |
Everything here is from Anthropic's What's new in Claude Fable 5.1 page and its pricing page, read on 16 September 2026. Apart from the model ID, the two highlighted rows are the only ones that differ.
Apart from the name, two rows differ. They are the whole story.
The real difference is the door
Mythos 5.1 is offered only to approved customers in Project Glasswing. Anthropic describes Glasswing as an effort to secure critical software for the AI era. Its launch partners include AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks, and it extended access to more than 40 other organisations that build or maintain critical software infrastructure.
For access, Anthropic’s docs point you at your Anthropic, AWS or Google Cloud account team. Open-source maintainers have a separate route through the Claude for Open Source programme.
If none of that describes you, the comparison is over. You cannot call Mythos 5.1, and nothing in its documentation suggests you are missing capability by using Fable 5.1 instead.
The one API difference
There is exactly one behavioural difference in the docs, and it favours Mythos for anyone whose code rewrites history.
On Fable 5.1, modifying anything that comes before a thinking block, whether the system prompt, the tools array or an earlier message, results in an error on the next request, or in that block being dropped if you opt in. It is one of the three breaking changes I went through in Fable 5.1 against Fable 5.
Anthropic’s docs say it directly: Claude Mythos 5.1 doesn’t run this check.
Forced tool use is documented as unsupported on both, so that breaking change applies either way. The one-way rule on thinking blocks, where no earlier Claude model can read them, is documented for Fable 5.1; the docs do not spell it out for Mythos 5.1, so test rather than assume.
What they share beyond the specs
Both carry 30-day data retention and are not available under zero data retention unless Anthropic expressly authorises it. Both are what Anthropic calls Covered Models, like Fable 5 and Mythos 5 before them.
Both also mark their output. Text generated by either carries Anthropic’s statistical text watermark on every platform where the model is available, and supported image, video and audio files come with signed C2PA Content Credentials when retrieved through the Files API.
So which one?
If you are not an approved Glasswing participant: Fable 5.1. It is the one you can use, and the docs give you no capability reason to want the other.
If you are: pick by the work the programme exists for, and treat Fable 5.1 as the compatibility target for anything that also runs outside it.
If what you are really deciding is whether a Fable-class model is worth paying for at all, that is a different question, and the more useful comparison is Fable 5.1 against Opus 5.
Questions people asked
Is Claude Mythos 5.1 better than Fable 5.1?
Not in capability, according to Anthropic, whose docs describe Mythos 5.1 as having the same capabilities as Fable 5.1. They share context window, max output, thinking, tokenizer and price. What differs is access and one API check.
How do I get access to Claude Mythos 5.1?
Through Project Glasswing only, by contacting your Anthropic, AWS or Google Cloud account team. Open-source maintainers can apply through the Claude for Open Source programme.
What is the API difference between Mythos 5.1 and Fable 5.1?
The history-editing check. Fable 5.1 can reject a request when anything before a thinking block was modified; Mythos 5.1 does not run that check. Forced tool use is unsupported on both.
Should I use Mythos 5.1 or Fable 5.1?
Fable 5.1, unless you are an approved Glasswing participant. If you are, still test integrations against Fable 5.1, where the history check applies.
If you are working out what a restricted model means for a product roadmap, the contact page is the fastest route to me.
