Anthropic's Claude Fable 5.2 Takes on OpenAI's GPT-6 Astra at a Lower Price

Admin


Anthropic debuted two new models on Monday, November 9 — Claude Fable 5.2 and Mythos 5.2 — its answer to OpenAI's GPT-6 Astra, which launched in early September. The pitch is straightforward: higher performance at a lower price. The new models are 20 percent cheaper than Astra for ordinary input and output tasks, at $8 per million input tokens and $40 per million output tokens, compared with $10 and $50 for Astra. For prompts loaded with large amounts of context, such as documents, the gap is far wider: Anthropic's models cost 20 cents per million cached tokens, against one dollar for Astra — an 80 percent discount.

The launch is critical for Anthropic. The company is seeking to re-accelerate revenue growth after a third-quarter slowdown revealed in its latest S-1 filing with the SEC, and the new models could be its last big releases before its initial public offering. Pricing is the battlefield: professionals have shown they will switch models quickly on cost, as long as performance holds up, and Anthropic is betting that a cheaper frontier model wins back users who drifted to OpenAI's Codex, which gained ground in recent months partly on lower pricing.

Anthropic's performance claims are bold. The company said the new models "helped our engineers complete otherwise-infeasible software migrations, designed three working electronic devices from scratch, solved two open problems in the physics of black holes, made substantial progress on robotics evaluations, and formalized the proof of Fermat's Last Theorem in 30 hours." It also said the models perform better than competitors on coding evaluation benchmarks, and that over 90 percent of the company's own code is now written by AI. Those are the company's own claims from pre-release testing, not independent results — but they signal where Anthropic believes the frontier has moved: from answering questions to doing long, open-ended technical work.

Anthropic positions Fable 5.2 and Mythos 5.2 for "the most ambitious, open-ended technical work," competing with Astra at the frontier. For everyday tasks, it still recommends its more cost-effective Opus 5.5 — a model that is more expensive than its direct OpenAI rival, GPT-6.1 Sol, though Anthropic argues it can still be the better deal if it produces a high-quality result more often. The message is that the model market now segments by job: the frontier model for the hardest work, the cheaper model for the routine — and the winner in each segment is decided as much by price as by capability.

Why it matters

This launch turns the AI model race into a price war at the top end, and that changes who benefits. For two years the frontier was defined by capability alone — whoever built the smartest model won. Now the two leading labs are competing on dollars per million tokens, with Anthropic undercutting OpenAI by a fifth on standard tasks and by four-fifths on cached context. For the companies building products on these models, that is real money: at the scale of millions of requests, a 20 percent price cut is the difference between a feature that is profitable and one that is not.

It also matters because of the timing. Anthropic is heading toward an IPO with a third-quarter slowdown to explain, and its answer is not a new product line but a cheaper, stronger core model. If Fable 5.2 and Mythos 5.2 live up to their claims in real-world use — the migrations, the devices, the proofs — Anthropic can argue that it, not OpenAI, sets the frontier. If the claims do not hold up outside the lab, the price cut looks like a discount on a second-place model, and the IPO story gets harder. Either way, the next few months of independent testing will decide more than any launch announcement can.

Source: Fortune — Anthropic debuts Claude Fable, Mythos 5.2 models, priced 20% below OpenAI's GPT-6 Astra (2026-11-09). Summary rewritten by VCTRIA.

Post a Comment