Claude Fable 5.1 Released Sep 1 2026: Cache Cut 75%

Quick Answer

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026 — Fable 5.1 priced at $10 per million input tokens and $50 per million output tokens, with cache reads cut 75% to $0.25 per million tokens. Mythos 5.1 remains restricted to vetted Glasswing coalition partners in cybersecurity and biology research. For typical workloads, the cache-read cut reduces total cost by about 25% relative to Fable 5; for highly agentic coding runs, the saving reaches roughly 45%.

Last verified: Sep 16, 2026.

At a glance

• Release date: September 1, 2026
• Models: Claude Fable 5.1 (general availability), Claude Mythos 5.1 (Glasswing only)
• Fable 5.1 input: $10 / MTok
• Fable 5.1 output: $50 / MTok
• Cache reads: $0.25 / MTok (down 75%)
• Typical workload savings: ~25%
• Agentic workload savings: ~45%
• Platforms: Anthropic API, AWS Bedrock, Google Cloud Vertex, Microsoft Azure

Why the cache read cut matters

Cache reads are the dominant cost driver for long-running agent workloads. In a retrieval-augmented generation pipeline, the same prompt prefix (system instructions, tool definitions, retrieved context) gets re-billed every turn without caching. Anthropic's prior pricing treated cache reads as roughly 25% the cost of fresh input tokens; the new pricing drops that to 2.5% — a 75% reduction. For an agent that runs hundreds of turns with a multi-thousand-token system prompt, the cumulative saving is substantial (Anthropic, September 1, 2026).

Anthropic framed the move as a direct response to customer feedback on cost predictability. "Alongside its increased capabilities, Fable 5.1 takes important steps towards addressing the feedback we've received from customers on price, data retention, and safeguards," Anthropic said in the announcement. The pricing change applies wherever usage is billed by token, including the Anthropic API, AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure AI Foundry (Anthropic, September 1, 2026).

What is unchanged

The base input/output pricing for Fable 5.1 is the same as Fable 5. $10/MTok input and $50/MTok output are unchanged; the savings come exclusively from cache reads. The release notes clarify that pricing for Fable 5.1 "is otherwise the same as Fable 5's" (Anthropic Fable 5.1 page, September 1, 2026).

For workloads that do not benefit from prompt caching — short, one-shot generations where the system prompt is small relative to the output — the price change is irrelevant. For workloads dominated by long-context agents, document Q&A with large retrievals, and multi-turn coding assistants, the savings are the headline.

Capabilities: coding, knowledge work, and research

Fable 5.1 is positioned as Anthropic's strongest model for coding and knowledge work, with research capabilities that the company describes as "an early glimpse of how AI models will contribute to scientific progress." The release emphasizes multi-step work, complex agentic flows, and the kind of long-horizon reasoning that benefits from prompt caching. Anthropic published supporting benchmarks in the announcement post; the model's headline positioning is "world's most advanced models for coding and knowledge work" (Anthropic, September 1, 2026).

For enterprise buyers who already standardized on Fable 5, the upgrade path is straightforward — the API model identifier `claude-fable-5-1` works with the same tooling, and prompt caches built against Fable 5 are reusable on Fable 5.1 with the new pricing applying automatically.

Mythos 5.1 and the Glasswing pattern

Mythos 5.1 is the higher-capability sibling model, and it remains restricted. Mythos 5 first launched in June 2026 with the Glasswing coalition — a vetted set of enterprise partners working in cybersecurity and biology research — because of the model's advanced cyber and biological capabilities. The September 1 update extends the same access pattern to Mythos 5.1 (Anthropic Mythos 5 page, June 9, 2026; Anthropic, September 1, 2026).

The Glasswing pattern mirrors OpenAI's Daybreak coalition for GPT-6 Astra (rolling out today, September 16, 2026). Both vendors are responding to the same pressure: how to deploy frontier-capability models to high-value enterprise users without compromising on the safety commitments that justified broad availability. The Glasswing/Daybreak model gives vendors a contractual envelope to work within while keeping the most capable models out of mass-market general availability.

What enterprise buyers should do

For existing Fable 5 workloads, the upgrade is mechanical — switch the model identifier and re-measure cost. Expect a 20-45% cost reduction on long-context agent workloads, depending on cache-hit ratio. For new workloads that would have been uneconomic at Fable 5 pricing (large document Q&A with 50k+ token context windows, multi-hour coding sessions), Fable 5.1 may now be cost-justifiable (Anthropic Fable 5.1 announcement, September 1, 2026).

For competitive evaluations against GPT-6 Astra, the comparison turns on workload type rather than raw capability: Fable 5.1 wins on cost for cache-heavy workloads; Astra wins on raw reasoning and the broadest tool ecosystem. Plan a 2-4 week evaluation window against representative production traffic before retiring either model.

What to watch next

Three near-term datapoints. First, Anthropic's Sonnet 5 family — currently priced at $2/$10 per million tokens — is the obvious candidate for a similar cache-read cut in a future release. Second, the AWS Bedrock and Azure Foundry pricing for Fable 5.1 should match the Anthropic direct pricing subject to provider terms; deviations will be a procurement red flag. Third, Mythos 5.1 access expansion to additional Glasswing partners will signal how Anthropic is calibrating cyber-biological risk tolerance.