Compare · ModelsLive · 2 picked · head to head

Grok 3 Mini vs o1-preview

Side by side · benchmarks, pricing, and signals you can act on.

Winner summary

Grok 3 Mini wins 3 of 5 shared benchmarks. Leads in knowledge · math.

Category leads
reasoning·o1-previewknowledge·Grok 3 Minimath·Grok 3 Minicoding·o1-preview
Hype vs Reality
Grok 3 Mini
#110 by perf·no signal
QUIET
o1-preview
#136 by perf·no signal
QUIET
Best value
Grok 3 Mini
116.5 pts/$
$0.40/M
o1-preview
no price
Vendor risk
xAI logo
xAI
$250.0B·Tier 1
Medium risk
OpenAI logo
OpenAI
$840.0B·Tier 1
Medium risk
Head to head
Grok 3 Minio1-preview
ARC-AGI
o1-preview leads by +1.5
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Grok 3 Mini
16.5
o1-preview
18.0
GPQA diamond
Grok 3 Mini leads by +34.6
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Grok 3 Mini
68.3
o1-preview
33.8
MATH level 5
Grok 3 Mini leads by +9.3
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Grok 3 Mini
90.9
o1-preview
81.7
OTIS Mock AIME 2024-2025
Grok 3 Mini leads by +46.7
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Grok 3 Mini
77.8
o1-preview
31.0
WeirdML
o1-preview leads by +5.0
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Grok 3 Mini
42.6
o1-preview
47.6
Full benchmark table
BenchmarkGrok 3 Minio1-preview
ARC-AGI
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
16.518.0
GPQA diamond
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
68.333.8
MATH level 5
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
90.981.7
OTIS Mock AIME 2024-2025
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
77.831.0
WeirdML
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
42.647.6
Pricing · per 1M tokens · projected $/mo at 10M tokens
ModelInputOutputContextProjected $/mo
xAI logoGrok 3 Mini$0.30$0.50131K tokens (~66 books)$3.50
OpenAI logoo1-preview