Benchmark family

FrontierMath

A family groups the versions of one benchmark. It is not evaluated and has no applicable models of its own; each of its 6 definitions is resolved separately.

MathematicsResearch mathematicsFormal reasoningMathematics

Definitions

Compare the newest 4 →

Newest first, following the recorded successor chain. Each definition is resolved against models separately.

BenchmarkFrontierMath Erdős2026-09
ActiveNewest in family
Scale
68 conjectures
Primary metric
Formal Proof Acceptance
Applicable models
0
BenchmarkFrontierMath Tier 4v2 (2026-06-12)
Active
Successor of FrontierMath Tier 4
Scale
43 problems
Primary metric
Accuracy
Applicable models
13
BenchmarkFrontierMath Tier 4v1 (old)
Superseded
Scale
50 problems
Primary metric
Accuracy
Applicable models
13
BenchmarkFrontierMath Tiers 1–3v2 (2026-06-12)
Active
Successor of FrontierMath Tiers 1–3
Scale
295 problems
Primary metric
Accuracy
Applicable models
13
BenchmarkFrontierMath Tiers 1–3v1 (old)
Superseded
Scale
300 problems
Primary metric
Accuracy
Applicable models
13
BenchmarkFrontierMath Tiers 1–4
Active
Scale
Not recorded
Primary metric
Not recorded
Applicable models
13

Family methodology

No family-level methodology recorded.

Across definitions

Tasks: None recorded
Sources: None recorded at family level
Last reviewed: 2026-09-22

Applicability does not mean evaluated; benchmark result/model score data is outside this UI contract.