← All benchmarks

OEIS Open

Math unit: % independent 5 models scored not in composite

Open problems from the Online Encyclopedia of Integer Sequences: find the rule behind a sequence that no one has formally characterised yet.

What it measures

Open-ended mathematical discovery rather than problem solving with a known answer.

How to read it

Extremely few models evaluated and scores are near zero across the board.

Source

Epoch AI Benchmarking Hub

Full ranking

#ModelLabScoreMeasuredMethodSource
1 Claude Fable 5 Anthropic 44.0% 2026-08-11 independent Epoch AI Benchmarking Hub
2 GPT-5.6 Sol OpenAI 43.0% 2026-08-11 independent Epoch AI Benchmarking Hub
3 Claude Opus 4.8 Anthropic 39.0% 2026-08-11 independent Epoch AI Benchmarking Hub
4 GPT-5.5 OpenAI 36.0% 2026-08-11 independent Epoch AI Benchmarking Hub
5 Gemini 3.5 Flash Google DeepMind 29.0% 2026-08-11 independent Epoch AI Benchmarking Hub