Anthropic Launches Claude Fable 5.2 With 20% Price Cut Ahead of IPO

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, the company’s first frontier model launch since CEO Dario Amodei called for a slowdown in AI development in September. The two models share the same architecture — Fable carries strong cybersecurity and biology safeguards for broad availability, while Mythos strips those safeguards for vetted organizations in cybersecurity and life sciences research.

“We called for pacing, not pausing,” Dianne Penn, Anthropic’s head of product management for research, told.¹ The models generate outputs up to 30% faster than their 5.1 predecessors and communicate “more clearly” with users, according to Reuters.

During prerelease testing, Fable 5.2 helped Anthropic engineers complete “otherwise‑infeasible software migrations,” designed three working electronic devices from scratch, solved two open problems in black hole physics, and predicted 28 of 34 capsid protein structures against experimental references, according to Anthropic.

The new models cost $8 per million input tokens and $40 per million output tokens — a 20% cut from Fable 5.1’s $10/$50 rates and also 20% below GPT‑6 Astra’s standard pricing, VentureBeat reported. An optional Fast mode runs at $16/$80 for up to 2.5x faster token generation. Batch API pricing halves those rates to $4/$20 — same as standard interactive Opus 5.5.

The per‑token numbers tell one story, but the per‑task economics are where things get interesting. On Cognition’s FrontierCode 2.0 test, Fable 5.2 at High effort averaged $6.72 per task — almost exactly what GPT‑6 Astra spent at Max ($6.68) but with a higher score (49.1% vs 42.8%). Turning Fable up to Xhigh bought the top score but raised the average bill to $14.19, per Anthropic’s system card.

Anthropic published a comparison table showing Fable 5.2 or Mythos 5.2 leading on 13 of 15 reported benchmarks against GPT‑6 Astra and Google’s Gemini 4 Carbon, according to Anthropic’s launch page. The coding lead is the clearest: Fable scored 49.1% on FrontierCode 2.0 (which tests whether a patch is good enough to merge into a real open‑source repo), compared to GPT‑6 Astra at 42.8% and Gemini 4 Carbon at 41.1%. On FrontierSWE v2, Fable hit 69.9% versus Astra’s 65.5%.

Carbon took one win on AutomationBench (53.9% to Fable’s 51.4%), and Opus 5.5 edged Fable on AA‑Briefcase’s head‑to‑head ranking — but the overall picture shows Fable 5.2 as the strongest all‑around frontier model available today.

In a striking demonstration of scale, Anthropic deployed 1,000 Fable 5.2 agents that formalized the proof of Fermat’s Last Theorem in 30 hours. An equivalent team of Opus agents completed roughly 27% of the work.

The 20% price cut matters, but Anthropic’s strategic shift is bigger than a discount. The company is positioning Fable 5.2 as a model that can act as an orchestrator — delegating simpler tasks to smaller, cheaper models while reserving its own compute for the hardest problems. Anthropic recommends developers start with Opus 5.5 for most tasks and escalate only the difficult ones to Fable.

Early users are seeing the difference. Vercel engineer Joe Haddad told Anthropic that projects his team “had spent weeks pushing along with other models were finished and approved in a few days” with Fable 5.2. Datadog senior staff engineer Matt Lee compared the model to an “incident commander” that could send faster models to gather information while it coordinated the work, according to testimonials published by Anthropic.

The cache pricing also got a boost. Cache reads now cost $0.20 per million tokens across Fable, Mythos, and Opus 5.5 — a significant reduction for teams reusing stable prompts, which is common in agentic workflows, per.³

The launch arrives at a critical moment. Anthropic’s IPO could happen as soon as next week, with the company targeting a valuation of $1 trillion to $1.5 trillion — down from the $2 trillion figure floated in recent months, CNBC reported. The company disclosed $16.3 billion in revenue and an $80.7 billion net loss for the first half of 2026 in its S‑1 filing last month.

The timing is delicate. Anthropic must convince public‑market investors that its safety‑first approach is a competitive advantage, not a brake on growth — all while OpenAI’s $11 billion funding round at a $1.25 trillion valuation and Google’s Gemini 4 Carbon launch keep the competitive heat high. The model launch serves as a proof point that Anthropic can still ship frontier capabilities on schedule.

Anthropic is positioning Fable 5.2 as a premium model for “power users or power use cases,” as Penn told.¹ For routine coding and knowledge work, the company still recommends Opus 5.5 as the daily driver, with Fable reserved for the hardest problems. This tiered approach — Haiku for speed, Sonnet for balance, Opus for daily work, Fable for frontier tasks, Mythos for specialized research — gives Anthropic the deepest model lineup of any AI lab.

The competitive landscape is crowded. Google’s Gemini 4 Carbon remains in limited release. Meta’s Muse Wonder, launched October 12, offers a lower‑priced option for sustained repository work. And DeepSeek’s V4 Pro 0813 costs just $1.32/$3.96 per million tokens, a fraction of Fable’s rates — though as VentureBeat noted, the comparison is incomplete without measuring per‑task spend on real workloads.