Full story
Fable 5.1 price cuts
Anthropic launched Claude Fable 5.1 and Claude Mythos 5.1 on Tuesday, positioning the pair as its latest models for coding and knowledge work while cutting cache read pricing.
“cache reads are now only $0.25 per million, down 75%”
The company said cache reads are now only $0.25 per million tokens, down 75%, while pricing for input/output tokens remains unchanged at $10/$50 per million input/output tokens.
In Anthropic’s benchmark results, Fable 5.1 scored 52.6% on Terminal-Bench-Science 0.1 versus Fable 5’s 24.7%, and the release also reported that on Terminal-Bench 4.0 Fable 5.1 scored 55.8% while Mythos 5.1 reached 60.9%.
Anthropic also described the two models as the same model under the hood but with different safeguards, with Fable 5.1 generally available and Mythos 5.1 restricted to Anthropic’s trusted access program for cybersecurity and the life sciences.
The New Stack framed the update as “a bit cheaper, a bit smarter,” while TechCrunch described it as “cheaper, less restrictive,” tying the changes to reduced token cost and false-positive restrictions from safeguards.
Safeguards, vulnerabilities, watermarks
Anthropic said it updated the cybersecurity and biology safeguards so they intervene less often on benign requests, with cyber safeguards now intervening about 60% less per session and biology safeguards intervening 85% less often on harmless questions.
The company also said the model is now allowed to identify vulnerabilities in code, while penetration testing, exploit generation, and scanning binaries for vulnerabilities remain off the table and are redirected to its Opus models.
On the privacy side, Anthropic said it is rolling out Enterprise Frontier Safeguards (EFS), which it described as storing data in cloud infrastructure controlled entirely by the customer, not Anthropic.
For content provenance, Anthropic shared a watermark technique for models released after August 2, 2026 and said it is opening a detection API in private preview “to eligible organizations as required under EU law (such as regulators, law enforcement, media, fact-checkers, independent researchers, educational organizations, and EU civil society groups).”
The Decoder quoted Anthropic’s benchmark framing as Terminal-Bench-Science 0.1 “a benchmark that tests whether an AI agent can carry out scientific research tasks inside a command-line environment,” while Mashable tied the release to “AI security worries mount.”
Access split and disputes
Access to Mythos 5.1 remains restricted, with The New Stack saying it is limited to Anthropic’s trusted access program for safeguards “specifically designed to support work in cybersecurity and the life sciences,” while Fable 5.1 is generally available.
“Fable 5.1 at max effort actually costs 20 percent more per task than Opus 5”
The Next Web reported that Mythos 5.1 is limited to US organizations through two programs, including the Cyber Verification Program and the Life Sciences Verification Program, and said Anthropic plans to expand access to international partners over time.
Pricing claims drew dispute, with The Decoder’s update from Artificial Analysis saying the cache read cut saves about $1.40 per task on agentic workloads but that at max effort Fable 5.1 “actually costs 20 percent more per task than Opus 5.”
The Decoder also reported that at max effort Fable 5.1 runs $3.76 per Intelligence Index task compared to Opus 5 at $2.34, and at extra-high effort Fable 5.1 scores 65 at $2.72 per task while still costing more than Opus 5.
Across coverage, The Next Web tied the watermark requirement to EU obligations by noting “Article 50requires providers to mark synthetic output in a machine-readable format and make it detectable,” while Cadena 3 Argentina quoted Anthropic’s assurance that “Anthropic has never trained on enterprise data without explicit permission and never will.”