Claude Fable 5.2 tops SimpleBench at 94.9%, half a point below the best human score, with Gemini 4 Carbon second at 92.6%

Claude Fable 5.2 is #1 on SimpleBench at 94.9%, half a point below the highest human score the benchmark records, 95.4%. SimpleBench's leaderboard data, read by ArtificialWatch at 18:50 UTC on November 10, added Fable 5.2 and both Gemini 4 models that day. @Mr_Salio flagged the update at 18:03 UTC.

THE TOP OF THE BOARD

SimpleBench scores are the average of five runs (AVG@5):

• 1st, Claude Fable 5.2: 94.9%
• 2nd, Gemini 4 Carbon: 92.6%
• 3rd, Gemini 4 Argon: 91.7%
• 4th, Claude Opus 5.5: 88.4%
• 5th, Claude Fable 5.1: 86.6%
• 6th, GPT-6 Astra Pro: 86.5%, OpenAI's highest entry
• 7th, GPT-6 Astra: 83.6%

WHAT THE NUMBERS SAY

• Fable 5.2 is 8.3 points above Fable 5.1, the model it replaced on November 9.
• SimpleBench's human baseline is 83.7%, from a sample of nine people with no special training, per the site. The top six models clear it; GPT-6 Astra, at 83.6%, falls 0.1 points short.
• The site's summary paragraph still names the original Claude Fable, at 81.9%, as the top model. The leaderboard table itself carries the November 10 entries.
• Also added November 10: Muse Ember at 72.7% (21st) and GPT-6.1 Terra at 72.2% (22nd).

HOW IT IS RUN

SimpleBench is a multiple-choice test of basic reasoning questions, from the AI Explained channel. The site lists temperature 0.7 and top-p 0.95 as its settings, and points to its technical report for how the human scores were measured.

WHAT WE DO NOT KNOW

Which reasoning-effort setting each model ran at, which the leaderboard does not list, and how much the 0.9-point gap between Carbon and Argon would move between runs. Today's top three were all added the same day.