[KongchangAI]
Model

CloudFable 5.2

A large language model that ties for first place with GPT-6.2 Astra on the Epoch AI comprehensive capability leaderboard.

Timeline (last 90 days)

Nov 21

A Bilibili uploader used three tests (Blender clock building, neon cube shooter, formula racing) to compare GPT-6.2 Astra and GPT-6.2 Sol, and also brought in CloudFable 5.2 and Opus 5.5 for a side-by-side comparison.

Unverified50%
Nov 21

In the Blender clock-building test, both Astra and Sol only produced a gearbox with a pendulum, lacking an escapement mechanism and unable to actually keep time; whereas both CloudFable and Opus produced a real escapement mechanism.

Unverified50%
Nov 21

Clock-building test costs: Astra $18, Sol less than $5; among historical results, Fable 5.2 was $36, Opus 5.5 was $42, GPT-6 Astra was $30, Kimi was $16, and Fable 5.1 was $125.

Unverified50%

All Facts (3)

Source Articles