Sección de la base de datos · 12 registros
Benchmarks
How capability and hardware are actually measured — and which yardsticks are already saturated.
Notas de campo
Los benchmarks son la forma en que esta industria lleva la puntuación, y su ciclo de vida es siempre el mismo: se presentan, se conquistan, se retiran. La pregunta interesante en cada registro no es el líder actual — es la velocidad a la que la prueba pasó de imposible a saturada.
Empiece por estos tres
ML
MLPerf
Los únicos números auditados de comparación directa en hardware de IA.
SB
SWE-bench Verified
Issues reales de GitHub — el referente de la era de los agentes de código.
AA
ARC-AGI
Diseñado para resistir la memorización; falló de todos modos. Luego reconstruido con más rigor.
Hardware & systems benchmarks
2
ML
MLPerf
The industry's referee for AI hardware — training and inference rounds.
TO
TOP500
The supercomputer league table since 1993 — now measuring the public sector's slice.
Model capability benchmarks
10
SB
SWE-bench Verified
Real GitHub issues, real repos — the coding-agent standard.
A(
AIME (Competition Math)
Olympiad-track problems — where reasoning models showed their jump.
AA
ARC-AGI
Chollet's abstraction puzzles — designed to resist everything but intelligence.
FR
FrontierMath
Research-grade mathematics — built unsolvable, already yielding.
GD
GPQA Diamond
Google-proof PhD science questions — the reasoning era's exam.
H&
HumanEval & GSM8K
The retired classics of coding and grade-school math.
HL
Humanity's Last Exam
2,500 expert questions with a deliberately final name.
L(
LMArena (Chatbot Arena)
Millions of blind human votes — preference as a leaderboard.
M&
MMLU & MMLU-Pro
The default knowledge exam of 2020–24 — now saturated.
MT
MTEB
The embedding leaderboard behind every retrieval stack.