HomeFeaturesPricingBlogFAQContact
← All articles

Claude Fable 5 Pricing: $10 Per Million Tokens and Leaving Subscriptions on June 23

Key takeaways
  • Anthropic released Claude Fable 5 on June 9, 2026 — a Mythos-class model scoring 95% on SWE-bench Verified and 80% on SWE-bench Pro, the strongest results Anthropic has publicly shipped.
  • Fable 5 is included at no extra cost on Pro, Max, Team, and Enterprise subscription plans through June 22, 2026, but will be removed from those plans starting June 23.
  • The standard Claude Fable 5 pricing is $10 per million input tokens and $50 per million output tokens — exactly twice the price of Claude Opus 4.8.
  • API and consumption-based Enterprise access continues indefinitely, while subscription users will fall back to Opus 4.8 as the top available model after the June 22 cutoff.

Anthropic released Claude Fable 5 on June 9, 2026, giving the public its first access to a Mythos-class model. The Claude Fable 5 pricing is $10 per million input tokens and $50 per million output tokens. But the more consequential number is June 23 — the date the model disappears from flat-rate subscription plans.

What Claude Fable 5 Is and Why It Is Different

Fable 5 is the publicly accessible branch of Anthropic's Mythos model family. The full Claude Mythos 5 is restricted to Glasswing security partners and a narrow cohort of biology researchers. Fable 5 is what Anthropic considers safe for general use: the same Mythos-class architecture with a safeguard layer that reroutes under 5% of sessions to Claude Opus 4.8 when queries hit specific sensitivity thresholds.

The performance numbers are meaningfully ahead of anything Anthropic has previously released without partner clearance. According to Anthropic's model card published on the Claude API documentation page, Fable 5 scores 95% on SWE-bench Verified and 80% on SWE-bench Pro. Those are the highest publicly reported results from any major lab on these software-engineering benchmarks as of June 2026. Anthropic also reports exceptional results on knowledge work, vision tasks, and scientific research evaluations.

The safeguard caveat is worth noting for cost planning: the under-5% of sessions that are rerouted to Opus 4.8 will be charged at Opus 4.8 rates, slightly reducing your effective average cost per token compared to the published Fable 5 rate.

The June 22 Cutoff for Subscription Users

Through June 22, Claude Fable 5 is included at no extra cost on Pro, Max, Team, and seat-based Enterprise subscription plans. Starting June 23, it is removed. Subscription users revert to Claude Opus 4.8 as their top available model.

For general writing, analysis, summarization, and everyday chat tasks, Opus 4.8 remains capable. The gap becomes material for heavy coding work, agentic multi-step tasks, and complex reasoning — the specific domains where Fable 5's benchmark lead over Opus 4.8 is largest.

Anthropic has not announced any subscription tier that will include Fable 5 as a permanent benefit. The 14-day free window reads as a structured preview: subscribers get to test the model before it moves fully into the consumption-based tier.

Claude Fable 5 Pricing Compared to Opus 4.8

The price doubles at every line item. Opus 4.8 costs $5 per million input tokens and $25 per million output tokens. Fable 5 is $10 input and $50 output. Whether the premium is justified depends entirely on the task.

For high-stakes software engineering, research, or agentic workflows where output quality directly affects downstream costs, a model scoring 95% on SWE-bench Verified may easily justify the 2x price if it reduces retries, errors, or human review. For routine tasks where Opus 4.8 already performs well, the extra cost is hard to defend.

The practical approach for most API users is to route by task: use Fable 5 for the work that genuinely needs its ceiling, and Opus 4.8 for everything else. June 2026 has made that kind of per-task model selection the main lever for controlling AI costs.

What the Launch Signals About Subscription vs API Access

The Fable 5 launch continues a pattern visible throughout the first half of 2026. Frontier models are increasingly API-first. Subscriptions get time-limited previews at most. GitHub Copilot moved to usage-based billing on June 1. Anthropic is now separating its highest-capability model from its subscription tier two weeks after launch.

The direction is consistent: flat-rate subscriptions are being repositioned as entry-level access, while ongoing access to the most capable models requires direct API usage billed per token.

This is the premise behind bring-your-own-key tools like ByteChat — using your own Anthropic API key means Fable 5 is available at the published rate, with no subscription calendar determining which model you can reach.

The trade-off is real: per-token billing requires tracking consumption in a way a flat subscription does not. Heavy Fable 5 usage could add up quickly at $50 per million output tokens. But for users who need the model's best, API access is currently the only path that does not expire on June 23.

Frequently asked questions

What is Claude Fable 5 and when did it launch?

Claude Fable 5 is Anthropic's publicly available Mythos-class AI model, released on June 9, 2026. It scores 95% on SWE-bench Verified and 80% on SWE-bench Pro, making it the strongest model Anthropic has publicly shipped to date.

Will Claude Fable 5 stay available on Pro and Max subscriptions?

No. Fable 5 is included in Pro, Max, Team, and Enterprise subscription plans through June 22, 2026 only. Starting June 23, those plans revert to Claude Opus 4.8 as the top available model. Ongoing Fable 5 access requires API or consumption-based Enterprise usage.

How does Claude Fable 5 pricing compare to Claude Opus 4.8?

Fable 5 costs $10 per million input tokens and $50 per million output tokens — exactly double the Opus 4.8 rate of $5 input and $25 output. The gap is easiest to justify on demanding coding, agentic, and research tasks where Fable 5's benchmark advantage over Opus 4.8 translates into fewer retries and higher first-pass quality.

Getting the best AI increasingly means accepting a per-token bill rather than a flat monthly one.

Use Claude Fable 5 on your own Anthropic API key

ByteChat lets you bring your own Anthropic key and access Claude Fable 5 at Anthropic's published rate — no subscription deadline, no token markup.

Try ByteChat free →