Benchmark

ARC-AGI Benchmark

What is ARC-AGI?

Compute per task is reported alongside score by the maintainers and should not be dropped. Benchmark Radar tracks 3 reported scores for ARC-AGI, cited by 4…

Released: 2019-11-05
Evaluates: reasoning
Openness: unknown
Importer: Model reports
Reported scores: 3

Source provenance

Original evidence https://arcprize.org/

ARC-AGI links

Website https://arcprize.org/

ARC-AGI results and reported scores

Model reports

Model	Organization	Reported value	Reported	Evidence
Seed2.0 Lite	ByteDance	77.5	2026-02-14	evidence
Seed2.0 Mini	ByteDance	72.3	2026-02-14	evidence
Seed2.0 Pro	ByteDance	85.4	2026-02-14	evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.

Which sources cite ARC-AGI?

Seed2.0 (Pro / Lite / Mini) ByteDance, 2026-02-14
Grok 4 xAI, 2025-07-09
Gemini 2.5 Pro Google, 2025-06-17
o3 and o4-mini OpenAI, 2025-04-16