// AI NEWS · MODEL RELEASE

Anthropic Releases Claude Fable 5.1 and Mythos 5.1 With Cheaper Cache Reads and Stronger Coding

The September 1, 2026 update holds Fable 5.1 at the same per-token price, cuts cache-read costs sharply, and adds mid-conversation effort control. Mythos 5.1 stays gated to trusted programs.

2026-09-01 · by Moe Ameen

What happened

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, an update to the Fable 5 and Mythos 5 line the company first shipped on June 9, 2026. As with the earlier release, the two are the same underlying model at different safeguard levels: Fable 5.1 is generally available, while Mythos 5.1 — with restrictions relaxed for cybersecurity and life-sciences work — is limited to vetted organizations through Anthropic's trusted-access programs, including its Cyber Verification and Life Sciences Verification programs. Fable 5.1 is available on Claude.ai and through the API (model id `claude-fable-5-1`) on AWS, Google Cloud, and Microsoft Azure.

On price, Fable 5.1 holds at $10 per million input tokens and $50 per million output — unchanged from Fable 5 — but Anthropic cut cache-read costs by roughly 75% (to about $0.25 per million tokens) and says typical workloads land around 25% cheaper overall, rising toward 45% for highly agentic tasks. The context window stays at one million input tokens with up to 128,000 tokens of output per response.

Two capabilities are new. The first is mid-conversation effort control, which lets you dial how hard the model works on a given step up or down without starting a fresh session. The second is content provenance tracking, which tags outputs so downstream systems can verify where generated content originated — Anthropic frames it partly around EU AI Act compliance. On benchmarks, Anthropic reports gains over Fable 5 in coding and agentic computer use; for example, Terminal-Bench 4.0 scores of about 55.8% for Fable 5.1 and 60.9% for Mythos 5.1. Partners including Jane Street and Cognition said they were shifting workloads onto the model at launch.

Treat the exact benchmark figures and pricing as launch-day snapshots that can move, and verify current numbers on Anthropic's page before relying on them. One thing to keep straight: Fable 5.1 is a text-and-reasoning model. It writes, analyzes, codes, and reasons — it does not generate images, video, or audio.

Why it matters for creators

  • Cheaper by default: if you build on Claude, the same or stronger output now costs less. Cache reads are down about 75% and typical workloads roughly 25%, which helps most on long, repeated-context jobs like drafting a week of content from one source.
  • Better coding and agentic scores mean tools built on Claude — agents, automations, content pipelines — get more reliable under the hood without you changing anything.
  • Content provenance tagging signals where disclosure is heading: traceable, watermarked AI output is becoming a compliance expectation rather than a nice-to-have.
  • Mythos 5.1 staying gated to trusted programs is a reminder that the unrestricted tier is not for general creators — the model you can actually use is Fable 5.1.
  • It is a clean, high-search launch with a simple story (same price, cheaper, stronger), which makes it good material for a timely explainer while interest peaks.

How to act on this with Kompozy

The quiet win here is that you may already be getting this upgrade. Kompozy runs its text generation server-side on managed Claude models, so when Anthropic ships a cheaper, stronger version of the model, your captions, scripts, blogs, and newsletters inherit it without you touching a setting or rotating a key. A roughly 25% cost cut and better reasoning scores show up in your pipeline as better drafts at lower cost, not as a migration project. And on the Founding tier, where you can bring your own Anthropic key, the new cache-read pricing flows straight to your own bill.

The launch itself is also content. Drop the facts into Kompozy as a source and the engine fans one take into a package — a Blog Article on what changed for builders, a Carousel comparing Fable 5.1 to Fable 5 slide by slide, a Listicle Video of the five takeaways over a portrait clip, and platform-native Text Posts — then schedules and publishes the set across the eight social platforms plus your blog on autopilot. You ship the explainer during the peak-interest window instead of assembling six formats by hand after the moment has cooled.

Quick takeaways

  • Anthropic shipped Claude Fable 5.1 (general release) and Mythos 5.1 (trusted-access only) on September 1, 2026, updating the June 9 Fable 5 / Mythos 5 line.
  • Fable 5.1 holds Fable 5 pricing ($10/$50 per million tokens) but cuts cache reads ~75% and runs ~25% cheaper on typical workloads, up to ~45% on agentic tasks.
  • Context stays at 1M input tokens with up to 128K output; new additions are mid-conversation effort control and content provenance tracking.
  • Fable 5.1 is a text-and-reasoning model — it does not generate images, video, or audio.
  • Kompozy runs generation on managed Claude, so users inherit the cheaper, stronger model automatically and can news-jack the launch into a full content set.

Frequently asked questions

What are Claude Fable 5.1 and Mythos 5.1?

They are Anthropic's September 1, 2026 update to its Fable 5 / Mythos 5 model line. Fable 5.1 and Mythos 5.1 are the same underlying model at different safeguard levels: Fable 5.1 is generally available, while Mythos 5.1 relaxes restrictions for cybersecurity and life-sciences work and is limited to vetted organizations through trusted-access programs.

Did Claude Fable 5.1 get cheaper than Fable 5?

Per-token pricing is unchanged at $10 per million input tokens and $50 per million output. What dropped is cache reads, cut roughly 75% to about $0.25 per million tokens; Anthropic says typical workloads run around 25% cheaper overall and up to 45% cheaper for highly agentic tasks. Confirm current rates on Anthropic's page.

What is new in Fable 5.1 versus Fable 5?

Alongside reported gains in coding and agentic computer use, Fable 5.1 adds mid-conversation effort control (dial how hard the model works on a step without a new session) and content provenance tracking that tags outputs so their origin can be verified. Context stays at one million input tokens with up to 128,000 output tokens.

Can Claude Fable 5.1 generate images or video?

No. Fable 5.1 is a text and reasoning model — it writes, analyzes, codes, and reasons but produces no images, video, or audio. To turn its writing into published video, carousels, and posts, pair it with a content engine like Kompozy that renders the media and publishes across platforms.

Related news

← All AI news · Get started →