Claude Fable 5 vs GPT-5.5 Pro: Comparison on 12 Tasks via BotHub Aggregator

The authors of the Habr blog BotHub compared two top language models — Claude Fable 5 from Anthropic and GPT-5.5 Pro from OpenAI — on 12 practical tasks, testing both models through their own API aggregator of neural networks.

Why this particular pair is being compared

In the previous article, the authors already compared Claude Opus 4.8, GPT-5.5, and Gemini 3.1 Pro, but immediately clarified: GPT-5.5 Pro did not participate at that time because it was more correct to compare it against Claude Fable 5 — but at the time of publication, Fable 5 was unavailable. The situation changed when Anthropic unexpectedly returned Claude Fable 5 to public access.

• Claude Fable 5 — the first model of the new Mythos class, which ranks above Opus in Anthropic's hierarchy
• Anthropic unexpectedly returned Fable 5 to open access
• The comparison is based on 12 practical, everyday tasks, not formal benchmarks
• Testing was conducted through the BotHub API aggregator, not through the web interfaces of the models

How the testing methodology is structured

The authors deliberately refused ready-made benchmarks and marketing claims from manufacturers, betting on real everyday tasks. Testing went through the BotHub neural network aggregator — it works through APIs, which, according to the authors, eliminates the crutches and workarounds that implicitly help models in their branded web interfaces. The methodology also allows you to immediately see how much each task costs in money.

Cost is measured in BotHub's internal currency — CAPS, tied to the number of tokens spent. Typically, one ruble buys 4000 to 6500 CAPS depending on purchase volume; in this comparison, the authors used a rate of about 1 ruble = 6370 CAPS, since their plan changed. Both models belong to the premium segment, so the final prices for tasks turned out to be significantly higher than in the previous comparison of Opus 4.8, GPT-5.5, and Gemini 3.1 Pro.

The winner of each task is determined mainly subjectively, as before, with the authors honestly warning readers about this. All testing materials have been released into the public domain, so readers can form their own opinions and choose their favorite among the models.

What makes this comparison format interesting

The open publication of all testing materials is part of the methodology: instead of simply declaring a winner, they give readers the opportunity to assess the answers of both models to each of the 12 tasks themselves and form their own opinion, even if it differs from the authors' conclusions. This approach is especially valuable for the premium segment of models like Claude Fable 5 and GPT-5.5 Pro, where the cost of the request itself becomes part of the decision about choosing a model — not only the quality of the response, but its price in CAPS.

What this means

The return of Claude Fable 5 to public access for the first time allows a direct comparison of Anthropic's top Mythos-class model with OpenAI's top GPT-5.5 Pro version on equal API terms, rather than through manufacturers' marketing promises. For teams choosing a model for specific tasks, such practical comparisons with transparent token cost accounting are more valuable than formal benchmarks, which premium models typically pass almost equally well.

Frequently Asked Questions

What is the Mythos class of models at Anthropic?

Mythos is a new class of models from Anthropic that ranks above Opus in the company's hierarchy. Claude Fable 5 is the first model released in this class.

How did the comparison calculate the cost of requests to the models?

Cost was measured in the internal currency of the BotHub aggregator — CAPS, tied to the number of tokens spent. Typically, 1 ruble equals 4000-6500 CAPS depending on purchase volume; in this test, a rate of about 1 ruble = 6370 CAPS was used.