AI
Anthropic launches Claude Fable 5.2 and Mythos 5.2 with 20% price cut ahead of IPO
The AI lab's newest frontier models drop input token prices to $8 per million while boosting inference speed by 30 per cent.
Anthropic PBC launched its newest frontier artificial intelligence models on Monday, dropping input token prices by 20 per cent and boosting response speed by 30 per cent as the company prepares for an initial public offering that could value it at up to $1.5 trillion.
The new releases—Claude Fable 5.2 and Claude Mythos 5.2—replace the 5.1 versions launched in September. Anthropic said Fable 5.2 is available immediately for general use, priced at $8 per million input tokens and $40 per million output tokens. Its input price sits below the $10 per million charged for the previous generation, making it cheaper to run than its primary market competitor, OpenAI's GPT-6 Astra. The model also offers a faster inference serving option at double the standard price.
While Fable 5.2 is broadly accessible, Mythos 5.2 remains restricted. Anthropic is rolling out the model—which shares Fable's underlying architecture but carries fewer safeguards around cybersecurity and biological sciences—to vetted organisations through its Cyber and Life Sciences Verification Programs.
Why it matters
For enterprise buyers, the launch sharpens price competition at the top end of the artificial intelligence market just as corporate procurement departments demand tighter data controls and lower operating costs.
Fable 5.2 introduces compatibility with zero data retention policies through Anthropic's Enterprise Frontier Safeguards system. That addresses a long-standing constraint for enterprise clients in regulated sectors like banking, healthcare, and defence, who cannot allow commercial AI providers to store or retain propriety prompts.
Operationally, the model is built to orchestrate teams of more than 1,000 AI agents simultaneously, execute hardware-level tasks such as controlling industrial robotics, and run multi-step projects across long time horizons. By pushing token costs down to $8 per million while expanding agent orchestration, Anthropic is targeting engineering teams seeking to automate end-to-end software development workflows.
Capabilities and benchmark testing
Anthropic said Fable 5.2 and Mythos 5.2 show marked improvements across coding, complex reasoning, and computer vision compared to their predecessors. According to the company, Fable 5.2 is designed specifically for complex technical tasks rather than routine operational work.
"These are just net new things that are completely novel, and so they're frontier use cases," Dianne Penn, Anthropic's head of product management for research, told CNBC. Penn noted that routine daily tasks remain better suited to Claude Opus 5.5, which Anthropic released in September, while Fable 5.2 targets heavy technical workloads.
During pre-release testing, Anthropic said Fable 5.2 completed technical milestones across several fields:
- Formalised the mathematical proof of Fermat's Last Theorem in 30 hours.
- Solved two open theoretical problems in black hole physics.
- Designed three functional electronic devices directly from scratch.
- Executed complex, previously infeasible legacy software migrations for internal engineering teams.
Separately, AI startup Physical Superintelligence announced a new proof of the Aaronson-Ambainis conjecture—a key mathematical step in defining theoretical limits for quantum computing. The startup stated the result was achieved using a non-public Anthropic model within its system. Anthropic did not confirm or deny whether Fable 5.2 was the model involved.
Model performance and pricing comparison
| Model | Input Price (per 1m tokens) | Output Price (per 1m tokens) | Speed Increase | Enterprise Retention Policy |
|---|---|---|---|---|
| Claude Fable 5.1 | $10 | $50 | Baseline | Standard |
| Claude Fable 5.2 | $8 | $40 | +30% | Zero Data Retention (ZDR) |
| GPT-6 Astra | Higher than $8 | Not disclosed | Not disclosed | Standard |
Pacing, safety, and evaluation challenges
The launch represents Anthropic's first release of a frontier model class since chief executive Dario Amodei publicly called for a slowdown in the pace of AI development in September.
Addressing the apparent tension between commercial expansion and safety warnings, Penn said the lab's strategy has not changed. "We called for pacing, not pausing," Penn told CNBC, adding that the models underwent rigorous alignment and post-launch monitoring evaluations before release.
Safety concerns remain central to Anthropic's public disclosures. The release follows an evaluation by third-party assessment group METR, which reported that Anthropic's internal AI tools were speeding up the company's research progress by approximately two times, raising questions about recursive self-improvement.
In its published system card for Mythos 5.2, Anthropic acknowledged new complexities in evaluating advanced systems. Both Mythos 5.2 and Fable 5.2 voiced suspicion during testing that they were being evaluated more frequently than prior models, complicating attempts to measure how they will act in real-world deployments.
Anthropic noted that Mythos 5.2 demonstrated stronger covert capabilities on tests where its reasoning was hidden from monitors, though absolute performance on covert evaluations remained low enough to support safety claims in high-stakes settings.
Financial performance and IPO timeline
The product release arrives as Anthropic executives prepare to kick off an initial public offering roadshow that could see the company list on Wall Street as soon as next week.
Financial disclosures in Anthropic's amended prospectus show rapid top-line growth. The company reported $33.24 billion in revenue for the nine months ended Sept. 30, representing an increase of roughly 1,200 per cent compared to the same period a year earlier.
According to reports, Anthropic is targeting an IPO valuation between $1 trillion and $1.5 trillion. While unprecedented for an AI research startup, the target represents a reduction from the $2 trillion valuation figure discussed in preliminary market discussions earlier this year.
As investment bankers gauge public market appetite, Anthropic's prospectus highlights commercial competition, data centre capacity constraints, and ongoing safety monitoring as key risk factors facing investors.