Claude Fable 5 and Mythos 5 Pricing: Anthropic's New $10/$50 Top Tier
Anthropic launched Claude Fable 5 and limited-preview Claude Mythos 5 at $10 input and $50 output per million tokens, with 1M context and 128K max output.
By AI Pricing Guru Editorial Team
AI Pricing Guru articles are maintained by the editorial workflow behind the site: daily pricing snapshots, provider source checks, and review passes for model launches, subscription limits, and billing changes.
Anthropic just launched Claude Fable 5 and Claude Mythos 5, and this is not a normal Opus refresh.
Fable 5 is Anthropic’s most capable widely released model. Anthropic describes it as a Mythos-class model made safe for general use, with stronger performance than any Claude model it has previously made generally available. Mythos 5 is the same capability class with broader safeguards lifted for approved Project Glasswing customers.
The pricing is also a clean new top tier: $10.00 per million input tokens and $50.00 per million output tokens for both Fable 5 and Mythos 5.
For current live rates across the whole Claude lineup, see our Anthropic Claude API pricing page or model your own workload in the token cost calculator.
The Claude mobile app model selector is already showing Fable 5, with the label “Included until June 22.” That app inclusion signal is separate from API billing, but it confirms Anthropic is already putting Fable 5 in front of Claude users.
Claude Fable 5 API pricing
| Item | Claude Fable 5 price |
|---|---|
| Input tokens | $10.00 / 1M |
| Cache hits and refreshes | $1.00 / 1M |
| 5-minute cache writes | $12.50 / 1M |
| 1-hour cache writes | $20.00 / 1M |
| Output tokens | $50.00 / 1M |
| Context window | 1M tokens |
| Max output | 128K tokens |
| API model ID | claude-fable-5 |
Fable 5 is generally available on the Claude API, Claude Platform on AWS, Amazon Bedrock, Vertex AI, and Microsoft Foundry beginning June 9, 2026.
Anthropic’s docs say adaptive thinking is always on for Fable 5. Raw thinking is not returned; developers can request summarized thinking, and the API supports effort controls, task budgets, memory tool use, context editing, compaction, and vision.
Claude Mythos 5 pricing and access
| Item | Claude Mythos 5 |
|---|---|
| Input tokens | $10.00 / 1M |
| Cache hits and refreshes | $1.00 / 1M |
| 5-minute cache writes | $12.50 / 1M |
| 1-hour cache writes | $20.00 / 1M |
| Output tokens | $50.00 / 1M |
| Context window | 1M tokens |
| Max output | 128K tokens |
| API model ID | claude-mythos-5 |
| Availability | Limited Project Glasswing access |
Mythos 5 is not generally available. Anthropic says it is offered to approved customers through Project Glasswing, with access through Anthropic, AWS, or Google Cloud account teams.
This is why Mythos is going to be a hot search term quickly: it is the restricted model line people have been waiting on, and Fable 5 is now the generally available path into that capability class.
Fable 5 vs Opus 4.8 pricing
Fable 5 is exactly 2x Opus 4.8 on standard API token pricing.
| Model | Input | Cached input | Output | Availability |
|---|---|---|---|---|
| Claude Fable 5 | $10.00 | $1.00 | $50.00 | Generally available |
| Claude Mythos 5 | $10.00 | $1.00 | $50.00 | Limited preview |
| Claude Opus 4.8 | $5.00 | $0.50 | $25.00 | Generally available |
| Claude Sonnet 4.6 | $3.00 | $0.30 | $15.00 | Generally available |
| Claude Haiku 4.5 | $1.00 | $0.10 | $5.00 | Generally available |
That puts Fable 5 between regular Opus pricing and the restricted Mythos capability narrative. It is not cheap, but it is also not priced like an unreachable research-preview model. Anthropic says Fable 5 and Mythos 5 are less than half the price of Claude Mythos Preview.
Budget examples
| Workload | Fable 5 / Mythos 5 | Opus 4.8 |
|---|---|---|
| 100K input + 20K output | $2.00 | $1.00 |
| 1M input + 200K output | $20.00 | $10.00 |
| 10M input + 2M output | $200.00 | $100.00 |
The output side matters. A long agent run with large final deliverables can become expensive fast, especially if the workflow does not use prompt caching well.
What changed for developers
Fable 5 introduces a refusal and fallback behavior that developers need to handle.
Anthropic says Fable 5 includes safety classifiers. Some requests can be declined with stop_reason: “refusal” as a successful HTTP 200 response rather than an API error. Anthropic also supports fallback handling so refused requests can be retried on another Claude model, with fallback credit to avoid paying prompt-cache costs twice.
That means production routers should treat Fable 5 as a premium route with explicit fallback logic, not just a blind replacement for Opus 4.8.
Buyer advice
Use Fable 5 where the model’s extra capability can change the outcome: long-horizon coding agents, deep research, difficult document analysis, complex vision tasks, autonomous software work, and high-value workflows where a failed result costs more than the token bill.
Keep Opus 4.8 and Sonnet 4.6 in the router. Opus 4.8 is now the cheaper premium Claude option at half the Fable 5 price. Sonnet 4.6 remains the practical default for production workloads that do not need the new frontier tier.
For Mythos 5, the story is access rather than price. If you are not in Project Glasswing or a trusted access path, the practical model to evaluate now is Fable 5.
My read
Fable 5 is Anthropic turning the Mythos narrative into a generally available product lane.
The headline is not just “$10/$50 pricing.” It is that Anthropic now has a public Claude tier above Opus, with 1M context, 128K output, adaptive thinking always on, and enough app visibility that searches for Fable 5 and Mythos 5 are likely to spike within hours.
For buyers, the move is straightforward: benchmark Fable 5 immediately, route only the hardest work to it, keep Opus 4.8 for cheaper premium runs, and watch Mythos 5 access because that is where the highest-interest enterprise and security story will sit.
Sources: Anthropic launch post, Anthropic model introduction docs, and Anthropic pricing docs.