Tertius News is an AI-native newsroom: an AI model reads the linked source articles below and extracts what each outlet reported, so you can compare their coverage side by side. How this works →

Anthropic Unveils Claude Fable 5.1 and Mythos 5.1, Cuts Cache Read Prices by 75%

Anthropic has released two new Claude models, Fable 5.1 and Mythos 5.1, with Fable 5.1 generally available and Mythos 5.1 restricted to vetted users. The company cut cache read prices by 75%, introduced EU-mandated watermarking, and reported significant benchmark improvements.

By Tertius News AI Desk 2 distinct · 2 mastheads · 2 articles Version 1 Coverage Sep 1, 2026 Published 1h ago

Claude Fable 5.1 and Mythos 5.1 Launch

Anthropic has released two new Claude models: Fable 5.1 and Mythos 5.1. According to the company, the two models are the same at their core, differing only in their safeguard levels. Fable 5.1 is generally available today on all platforms, while Mythos 5.1 is limited to vetted users through two access programmes covering cybersecurity and the life sciences. The biology programme for Mythos 5.1 was built with the US government, as reported by The Next Web.

Anthropic describes Fable 5.1 as setting a new standard for coding, knowledge work, and long-running problem-solving tasks, claiming it is among the "world's most advanced models for coding and knowledge work." The company says Fable 5.1 achieves similar or better results than its predecessor at low or medium effort and shows "much higher performance" at higher effort tiers. In benchmark results cited by The Next Web, Fable 5.1 scored 52.6% on Terminal-Bench-Science 0.1 compared to 24.7% for Fable 5, and 55.8% on Terminal-Bench 4.0. MacRumors also reports the model outperforms Fable 5, Opus 5, and OpenAI's GPT–5.6 Sol across multiple benchmarks.

Pricing and Cost Reductions

A key selling point of the release is the cost reduction. Anthropic has cut cache read prices by 75%, bringing them to $0.25 per million tokens. According to The Next Web, this reduction lowers typical workload costs by about 25% and "agentic" workloads by up to 45%. MacRumors notes that base pricing remains the same, meaning the savings come entirely from the cache-read pricing change.

Safety Features and EU Compliance

Both models include improved safety measures. MacRumors reports that Claude Code users should see around 60% fewer cybersecurity false positives, and Fable 5.1 can now be used to discover software vulnerabilities, though not to develop exploits. The models also carry an invisible watermark on generated text, a requirement under the EU AI Act. The Next Web explains that Anthropic signed the EU's Code of Practice on transparency of AI-generated content in July, alongside 190 other signatories, and that the watermark applies to models released after 2 August 2026. The watermark is invisible without a detection tool and carries no information about the user.

Anthropic is rolling out a detection API in private preview for eligible organisations, including regulators, law enforcement, media, fact-checkers, independent researchers, educational bodies, and EU civil society groups, as required by EU law. The company says it plans to expand access over time but has not given a date.

The Next Web notes the European Union's Article 50 requires providers to mark synthetic output in a machine-readable format and make it detectable—a duty that has applied since 2 August 2026. The report also highlights that Mythos's access programmes are open only to a set of US organisations, two years after Claude arrived in the EU. European cyberdefenders and life scientists are not currently included, though Anthropic says it is working to include international partners.

How each outlet told it

A framing line is our reading of that outlet's own text — an interpretation, not a quotation and not a fact we assert. Check it against what the outlet published.

MacRumors

Framing: Focuses on the launch of new models, highlighting lower costs and fewer false positives.

as of Sep 1, 2026, 8:46 PM UTC

Facts Included:

It achieves similar or better results than Fable 5 at low or medium effort, and has much higher performance at higher effort tiers.
The API is available to eligible organizations as required by EU law.
Base pricing remains the same.
It is available today on all platforms.

Read the source article →

The Next Web

Framing: Highlights the release of new models and the 75% cache read price cut, focusing on pricing as the key selling point. — informative and business-focused

as of Sep 1, 2026, 9:47 PM UTC

Facts Included:

Anthropic has released two new Claude models: Fable 5.1 and Mythos 5.1, which are the same model with different safeguards.
Only one of them is generally available (Fable 5.1); Mythos 5.1 goes to vetted users through two access programmes covering cybersecurity and the life sciences.
The biology programme was built with the US government.
Cache read prices fall 75% to $0.25 per million tokens, cutting typical workload costs by about 25% and agentic ones by up to 45%.
Claude's outputs carry a watermark required by the EU AI Act.
Anthropic signed the EU’s Code of Practice on transparency of AI-generated content in July, alongside 190 other signatories.
The watermark applies to models released after 2 August 2026, is invisible without the detection tool, and carries nothing about the user.
The detection API is in private preview for eligible organisations, including regulators, law enforcement, media, fact-checkers, independent researchers, educational bodies and EU civil society groups.
Article 50 requires providers to mark synthetic output in a machine-readable format and make it detectable, and has applied since 2 August 2026.
Anthropic says it plans to expand access over time, without giving a date.
Mythos’s access programmes were built with the US government and are open only to a set of US organisations, two years after Claude arrived in the EU.
Anthropic says it is working to include international partners.
European cyberdefenders and life scientists are not in the programme today.

Read the source article →

What was claimed, and who backed it

AI-extracted; can misattribute a claim — see Methodology.

Each row is one claim, attributed to the outlet whose wording states it most clearly. Confidence rates how directly the source text states the claim — explicit and unhedged rates high; hedged, pieced-together, or internally inconsistent statements rate lower. It does not measure whether the claim is true. Status is Contested when two claims on this page negate each other; otherwise it counts the distinct outlets we found asserting that specific claim — so a single-source claim can still show high confidence, and a multi-source claim can show medium. Every one of those outlets is named beside the status, so you can check the count against the list. For claims extracted before we began storing that list, the row says so: it names the outlet the claim is quoted from and states that we have not recorded which outlets backed it. Outlets wrote at different times, so a figure that evolves — a casualty count, for example — can legitimately differ between rows; check the "as of" time next to each claim's source.

Claim	Confidence	Status
ClaimAnthropic has released two new Claude models: Fable 5.1 and Mythos 5.1, which are the same model with different safeguards.	ConfidenceHigh	StatusMulti-source (2 outlets · 2 distinct) via The Next Web as of Sep 1, 2026, 9:47 PM UTC, MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimOnly Fable 5.1 is generally available; Mythos 5.1 is limited to vetted users through two access programmes covering cybersecurity and the life sciences.	ConfidenceHigh	StatusMulti-source (2 outlets · 2 distinct) via The Next Web as of Sep 1, 2026, 9:47 PM UTC, MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimThe biology access programme for Mythos 5.1 was built with the US government.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimCache read prices for Claude fall 75% to $0.25 per million tokens, cutting typical workload costs by about 25% and agentic ones by up to 45%.	ConfidenceHigh	StatusMulti-source (2 outlets · 2 distinct) via The Next Web as of Sep 1, 2026, 9:47 PM UTC, MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimClaude's outputs carry a watermark required by the EU AI Act.	ConfidenceHigh	StatusMulti-source (2 outlets · 2 distinct) via The Next Web as of Sep 1, 2026, 9:47 PM UTC, MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimAnthropic signed the EU's Code of Practice on transparency of AI-generated content in July, alongside 190 other signatories.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimThe watermark applies to models released after 2 August 2026, is invisible without the detection tool, and carries nothing about the user.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimThe detection API is in private preview for eligible organisations, including regulators, law enforcement, media, fact-checkers, independent researchers, educational bodies and EU civil society groups.	ConfidenceHigh	StatusMulti-source (2 outlets · 2 distinct) via The Next Web as of Sep 1, 2026, 9:47 PM UTC, MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimArticle 50 requires providers to mark synthetic output in a machine-readable format and make it detectable, and has applied since 2 August 2026.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimAnthropic says it plans to expand access to the detection API over time, without giving a date.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimMythos's access programmes were built with the US government and are open only to a set of US organisations, two years after Claude arrived in the EU.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimAnthropic says it is working to include international partners in the access programmes.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimEuropean cyberdefenders and life scientists are not in the access programme today.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimFable 5.1 achieves similar or better results than Fable 5 at low or medium effort, and has much higher performance at higher effort tiers.	ConfidenceHigh	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimFable 5.1 outperforms Fable 5, Opus 5, and OpenAI's GPT–5.6 Sol across multiple benchmarks.	ConfidenceHigh	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimThe API is available to eligible organizations as required by EU law.	ConfidenceMedium	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimBase pricing for Claude Fable 5.1 remains the same as before.	ConfidenceHigh	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimClaude Fable 5.1 is available today on all platforms.	ConfidenceHigh	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimClaude Code users should see around 60 percent fewer cybersecurity false positives with Fable 5.1.	ConfidenceMedium	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimFable 5.1 can now be used to discover software vulnerabilities, but not develop exploits for them.	ConfidenceHigh	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC
ClaimFable 5.1 scores 52.6% on Terminal-Bench-Science 0.1 against 24.7% for Fable 5, and 55.8% on Terminal-Bench 4.0.	ConfidenceHigh	StatusSingle-source via The Next Web as of Sep 1, 2026, 9:47 PM UTC
ClaimAnthropic says Fable 5.1 and Mythos 5.1 are the world's most advanced models for coding and knowledge work.	ConfidenceHigh	StatusSingle-source via MacRumors as of Sep 1, 2026, 8:46 PM UTC

Transparency & Audit Trail

2 distinct · 2 mastheads · 2 articles consulted: MacRumors, The Next Web 22 claims extracted Version 1 Written 2026-09-01