Claude Fable 5.2 tops Artificial Analysis Intelligence Index and costs the most

Artificial Analysis says Claude Fable 5.2 scored 65 on its Intelligence Index. That is first out of 74 comparable models, against a median of 40. The reading is Artificial Analysis's own, taken after Anthropic released the model on 9 November 2026. On the same test suite, Fable 5.2 ran 44.1 minutes per task and cost $26.54. The medians across comparable models are 12.9 minutes and 72.8 million output tokens. Fable 5.2 produced 275 million.

The number and the index

The Intelligence Index combines nine benchmarks for agentic knowledge work, coding and scientific research. Artificial Analysis lists them as AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0 and FrontierCode 2.0. The rest are AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset) and Terminal-Bench-Science 0.1. The score reported is the average over those nine.

Fable 5.2's 65 is the highest score Artificial Analysis records. The second group clusters around 60 on the same index, and the median across every comparable model is 40. On that one measure, Anthropic's new top model sits well above every other model Artificial Analysis tracks.

What it costs to run

The intelligence number is one of several the dashboard reports. Three of them read the other way.

Fable 5.2 ran 44.1 minutes per task on the index. The median of the comparable models was 12.9 minutes. One task on Fable 5.2 therefore costs about 3.4 times as much wall-clock time as one on the median model.

It cost $26.54 per task. Artificial Analysis does not publish a median cost. Anthropic's own list price is $8 per million input tokens and $40 per million output tokens. The second surprise sits in the output volume. Fable 5.2 produced 275 million output tokens across the index, against a median of 72.8 million. The model reads about 3.8 times the output of the median model to reach its answer.

What Anthropic's own release says

Anthropic published Fable 5.2 and Mythos 5.2 on 9 November. The vendor page lists an input token price of $8 per million. It also claims a 30 percent speed increase on Fable 5.2 and zero-data-retention support. The release note says Fable 5.2 has a 1 million token context window.

Artificial Analysis is testing the "Max, Default Fallback" variant. The dashboard lists that configuration as the one Anthropic sets as a default.

A team that routes work to the strongest model should put a number against the extra it costs. On this benchmark set, the strongest model takes about 3.4 times the time of the median. It also reads about 3.8 times the output tokens. Those two gaps multiply at the API door.

For an agent that answers one query a second, the time number now caps throughput. The intelligence score sits beside that constraint, each measuring a different thing.

Source

Artificial Analysis: Claude Fable 5.2 model card