Claude Fable 5 is Anthropic's most powerful model that anyone can use, released on 9 June 2026. It sits in the new "Mythos-class" tier, handles up to 1 million tokens of context, and tops almost every coding and reasoning benchmark Anthropic tested it on. Pricing is $10 per million input tokens and $50 per million output tokens.
What is Claude Fable 5? (Quick Answer)
Claude Fable 5 is a frontier AI model from Anthropic, the company behind the Claude family. It belongs to a tier called "Mythos-class," which Anthropic places above its earlier Opus models. The short version: it is the strongest Claude model the public can access, with safety filters bolted on so it can be released widely.
Claude Fable 5 at a glance
| Feature | Claude Fable 5 | Why it matters |
|---|---|---|
| Release date | 9 June 2026 | Available the same day on the Claude API, AWS Bedrock, Vertex AI and Microsoft Foundry |
| Context window | 1 million input tokens | Can read an entire large codebase or a stack of long documents at once |
| Output limit | 128,000 tokens | Room for very long answers, full files, or detailed reports |
| SWE-bench Verified | 95.0% | Up from 88.6% on Opus 4.8 - a real jump in real-world coding |
| Pricing | $10 in / $50 out per million tokens | Less than half the price of the earlier Mythos Preview |
| Inputs | Text and vision | It can read images, charts and screenshots, not just text |
Why people are paying attention
The headline is coding. Stripe tested Fable 5 on a 50-million-line Ruby codebase and had it run a codebase-wide migration in a single day - work the company said would have taken a team more than two months by hand. On SWE-bench Verified, a standard test of fixing real software issues, it scored 95%. That is the kind of number that makes engineering teams rethink how they plan their week.
It is also more careful with its own work. At higher effort settings, Fable 5 reviews and checks what it produced before handing it back, which is part of why it can run long jobs without drifting off track. The longer and messier the task, the bigger its lead over older models.
The vision upgrade nobody expected
Fable 5 is now Anthropic's best model for anything involving images. It can pull exact figures out of a dense scientific chart, and it can rebuild a web app's source code from nothing but screenshots. The demo people keep sharing is a small one: earlier Claude models needed a pile of helper tools to play Pokémon FireRed. Fable 5 finished the game using only raw screenshots of the screen - no maps, no hints, no game-state feed. Silly on the surface, but it shows the model can look at a picture and figure out what to do next.
What the safety filters actually do
Here's the part that confuses people. Anthropic decided Fable 5 was too capable to release without guardrails, so it added a set of classifiers that watch for risky requests in three areas: cybersecurity, biology and chemistry, and attempts to copy the model's abilities. When one of those filters fires, your request quietly gets answered by Claude Opus 4.8 instead, and you are told it happened. Anthropic says this affects fewer than 5% of sessions. For the other 95%, you are talking to the full-strength model.
There is a sister model too, Claude Mythos 5 - the same brain with the cyber filters removed - but that one is locked to a small group of vetted cyber-defenders and, soon, biology researchers. Most users will never touch it.
What it costs and how to get it
At $10 per million input tokens and $50 per million output, Fable 5 is priced under half of the Mythos Preview that came before it. Developers can call it through the Claude API as claude-fable-5. For subscription users, Anthropic is rolling it out in stages because it expects heavy demand: from launch through 22 June it is included on Pro, Max, Team and seat-based Enterprise plans at no extra cost. On 23 June it moves to usage credits on those plans, and Anthropic says it intends to fold it back into standard plans once it has enough capacity.
Should you care if you're not a programmer?
Mostly yes, but with patience. The clearest wins right now are in software engineering, financial analysis, and research-heavy work. Early testers in legal and finance reported its output matching or beating their existing tools in blind reviews. If you write code, analyse documents, or do anything that involves long, multi-step reasoning, this is worth trying. If you just want a chatbot for everyday questions, Opus 4.8 or Sonnet will already do the job, and you may not notice the difference until tasks get genuinely hard.
The takeaway
Claude Fable 5 is a real step up, especially for coding and any job that runs for hours rather than seconds. The price drop and the wide launch mean a lot more people will get to try a frontier model that, six months ago, would have been locked behind a research program. Watch the 23 June pricing change if you are on a subscription plan, and keep in mind that roughly one in twenty requests on sensitive topics will be handled by Opus instead. For most people, most of the time, you will be using the best model Anthropic has ever shipped to the public.
Author: Dinesh Sharma is Founder & CEO of Hindustan Computer Institute and trains students in AI, cybersecurity, digital skills, and computer education.
Last Updated: June 2026
