Claude Fable 5.1: what it costs, what is free, and what the free plan actually gives you

Anthropic released Claude Fable 5.1 on 1 September 2026, alongside a second model called Mythos 5.1. If you are trying to work out whether you can use it without paying, the honest answer has two halves — and the second half is better news than you might expect.

What changed in 5.1

Anthropic positions Fable 5.1 and Mythos 5.1 as its most advanced models for coding and knowledge work. The release focused on two practical things: reducing token cost, and cutting down false-positive refusals — cases where the safety systems block a request that was fine all along.

It is described as better at long, complicated programming tasks — work spanning an entire application, such as a large refactor or a full code review — and at scientific work like designing experiments, simulating outcomes and reading dense diagrams.

The published benchmarks

Anthropic published these figures with the release. They are the company's own numbers, which is worth remembering when comparing against a competitor's own numbers.

• Terminal-Bench 4.0 — 55.8% for Fable 5.1, 60.9% for Mythos 5.1
• CursorBench 3.2.0 — 73.4%
• Humanity's Last Exam — 60.9% without tools, 65.0% with tools
• Terminal-Bench-Science 0.1 — 52.6%

What it costs

Anthropic publishes API pricing per million tokens:

• Input — $10 per million tokens
• Output — $50 per million tokens
• Cached reads — $0.25 per million tokens, a 75% reduction on the previous rate

What the free plan gives you

This is the part worth taking seriously rather than treating as a consolation prize. The Claude free plan includes Sonnet 5 as the default model and Haiku 4.5 as a faster option, on web, iOS and Android. No student status, no card, no region restriction.

For writing, summarising, everyday coding help and research, Sonnet 5 is a capable model. The case for Fable 5.1 is specifically long-running agentic work, large multi-file codebases and hard scientific reasoning. If that is not what you are doing, you may be shopping for capability you will not use.

A note on Mythos 5.1

Routes worth checking before you pay

Anthropic does not currently run a public student discount comparable to the offers below, so the honest advice is: use the free plan, and spend your student eligibility on the offers that do exist. These are the ones in our catalogue that overlap with the same kind of work.

The GitHub Student Developer Pack is the single highest-value item for a student developer, because one verification unlocks dozens of other offers. It includes GitHub Copilot at the student tier.

Worth being precise here, because round-ups routinely get it wrong: the pack includes Copilot Student, not Copilot Pro. You get unlimited code completions and an allowance of AI credits, with chat and agent usage limited. We checked this against GitHub's own page rather than repeating the common claim.

Should you upgrade to Pro?

The honest test is not whether Fable 5.1 is better than Sonnet 5 — it is. The test is whether the specific things it is better at are things you do. Anthropic was unusually clear about what those are, so you can check yourself against the list.

Upgrade if you regularly ask a model to work across an entire codebase rather than a single file: large refactors, full code reviews, migrations spanning dozens of files. That is the case 5.1 was built and benchmarked for, and it is where the difference is largest.

Upgrade if you run long agentic sessions where the model works for many steps without you correcting it each time. Stronger long-running behaviour compounds — a model that drifts on step twelve costs you the whole run.

Do not upgrade yet if your use is writing, summarising, research, or coding help one file at a time. Sonnet 5 on the free plan handles that, and you can always upgrade the month you need more.

If you are paying for the API instead

The pricing structure rewards a specific working style, and it is worth designing around rather than discovering on your bill.

• Output is five times the price of input. Asking for a concise answer is not just good style, it is the single biggest lever on cost.
• Cached input is forty times cheaper than fresh input. If you send the same codebase or document set repeatedly, structure your calls so that context is cached rather than re-sent.
• Adjustable thinking effort means you can spend more compute on hard problems and less on easy ones. Defaulting everything to maximum effort is the most common way to overspend.

Common questions

Can I use Claude Fable 5.1 for free anywhere?

Not through Anthropic. Any site offering free Fable 5.1 access is paying per token for your usage, which means the cost is being recovered some other way. The genuinely free option is the Claude free plan with Sonnet 5 and Haiku 4.5.

Is the free plan available in Bangladesh?

Claude's free plan is not restricted the way the US-only student offers are, so yes in the general case. Availability by country can change, so confirm on Anthropic's own site — but this is not one of the offers with a documented US-only restriction.

What is the difference between Fable and Mythos?

Mythos 5.1 is the same generation with fewer safeguards, aimed at cybersecurity and life-sciences work, and it scores higher on Terminal-Bench 4.0 — 60.9% against Fable's 55.8%. It is restricted to vetted US organisations through trusted access programmes, so for almost everyone reading this, Fable is the one that matters.

Does the student pack include Claude?

No. The GitHub Student Developer Pack includes GitHub Copilot at the student tier, not a Claude subscription. We mention the pack here because it covers similar coding work, not because it unlocks Anthropic's models.

Sources

Checked on 10 September 2026. Model pricing and plan availability change often — verify against Anthropic's own pages before relying on any figure here.

• Anthropic, 'Introducing Claude Fable 5.1 and Claude Mythos 5.1' — benchmarks and pricing
• Claude Help Centre, 'Claude Fable models on your plan' — plan availability
• TechCrunch, 1 September 2026 — cost and refusal-rate changes