Breaking

Anthropic Releases Claude Fable 5.1, Cutting Typical Token Costs by About 25%

September 1, 2026 at 15:17 EDT

Foundation Models
Software Dev & Coding
AI Agents

Anthropic on Tuesday released Claude Fable 5.1, a generally available successor to Claude Fable 5 that the company says is its most capable model for coding, knowledge work, and long-running agentic work. The same launch cuts cache-read prices by 75%, which Anthropic estimates will lower typical token-billed costs by about 25% and highly agentic workloads by up to about 45%.

September 1, 2026 · Anthropic

Claude Fable 5.1, the new public frontier

A Mythos-class upgrade over Fable 5: stronger long-horizon coding and research, the same $10 / $50 token rates, and cache reads cut to one-quarter of the prior price.

Cost of a Fable-class run

~25%

typical workload save

~45%

highly agentic save

75%

cheaper cache reads

$10 / $50

input / output per MTok

Agentic scientific research

Terminal-Bench-Science 0.1 · stacked blocks are proportional to score

52.6%

Fable 5.1

29.0%

Opus 5

24.7%

Fable 5

22.4%

GPT-5.6 Sol

Cache reads now cost 1×, not 4×

Same unit: dollars per million cached input tokens. Column height is the price.

$1.00

Fable 5 rate

$0.25

Fable 5.1

Long agent sessions that re-read a cached prefix now pay one-quarter as much on those tokens.

Where it gains

Partners report stronger unattended coding, a 52.6% science-bench score versus 24.7% for Fable 5, and cheaper Fable-class runs after the cache-read cut.

Where it still bites

Reviews can run slower, token use stays high, some answers overclaim evidence, and cyber or biology prompts may still fall back to Opus.

Who can use it

Paid Claude plans and the major clouds, not Free

Fable 5.1 is live on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, and Claude Platform on AWS. Claude Code needs version 2.1.250 or later.

Max / premium seats: up to 50% of weekly usagePro / standard Team: usage creditsMythos 5.1: trusted access only

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.