On November 9, 2026, independent benchmarking outfit Artificial Analysis updated its Intelligence Index (v5.0.1), and Anthropic's newly released Claude Fable 5.2 took the top spot with 65 points — five ahead of Google's Gemini 4 Carbon (60), six ahead of Anthropic's own Claude Opus 5.5 (59), and 16 above its predecessor Fable 5.1 (49). The gains came mostly from scientific reasoning, knowledge, and coding. It is the first independent third-party ranking for Fable 5.2.
How the 65 was earned
The Intelligence Index combines nine evaluations, including AA-Briefcase, Terminal-Bench 4.0, FrontierCode 2.0, and HLE-Diamond, covering agentic knowledge work, terminal coding, science, and knowledge Q&A. Fable 5.2 posted the highest score on five of the nine, and the top three entries on the board are all Fable 5.2 at different effort settings. On price, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens — 20% cheaper than Fable 5.1 — with a 1-million-token context window. One caveat: Artificial Analysis ran the tests with Anthropic's default fallback enabled, so requests that trip cybersecurity or biology safeguards are handed to another model, and that part of the score should be read with a discount.
The other side of the top spot is cost
For anyone choosing a model, the top ranking cuts both ways. On one side, the capability is genuinely a step up, especially for long-running research and engineering work. On the other, so is the cost: one full run of the index takes 45 minutes on average, more than three times the 12.9-minute median for models in the same price tier, the first answer token takes 171 seconds, and the whole evaluation generated 275 million output tokens. The top spot shows it is stronger on hard tasks; it does not mean it is worth using for everyday work, and its price sits in the top tier too. Other vendors are playing the same ranking game: Meta cited Artificial Analysis composite benchmark scores when it launched Muse Ember, writing value for money into the launch announcement.
Two things are worth watching next: whether other independent evaluations produce the same ranking, and how Fable 5.2 performs on real codebases, since the index runs standardized tasks. Anthropic positions Fable 5.2 for the hardest technical work and still recommends starting with Opus 5.5 for everyday coding. A top ranking is a signal, not a verdict — pick a model by what your own tasks look like.
