Compare · ModelsLive · 2 picked · head to head
Claude 3.5 Sonnet vs o1-preview
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
o1-preview wins on 6/9 benchmarks
o1-preview wins 6 of 9 shared benchmarks. Leads in arena · math · general.
Category leads
coding·Claude 3.5 Sonnetarena·o1-previewknowledge·Claude 3.5 Sonnetmath·o1-previewgeneral·o1-previewreasoning·o1-preview
Hype vs Reality
Attention vs performance
Claude 3.5 Sonnet
#203 by perf·no signal
o1-preview
#187 by perf·no signal
Vendor risk
Who is behind the model
Anthropic
$965.0B·Tier 1
OpenAI
$840.0B·Tier 1
Head to head
9 benchmarks · 2 models
Claude 3.5 Sonneto1-preview
Aider · Code Editing
Claude 3.5 Sonnet leads by +4.5
Claude 3.5 Sonnet
84.2
o1-preview
79.7
Chatbot Arena Elo · Overall
o1-preview leads by +14.5
Claude 3.5 Sonnet
1374.0
o1-preview
1388.5
Cybench
Claude 3.5 Sonnet leads by +7.5
Cybench · evaluates AI on real Capture-The-Flag cybersecurity challenges, testing vulnerability analysis, exploitation, and security reasoning.
Claude 3.5 Sonnet
17.5
o1-preview
10.0
GPQA diamond
Claude 3.5 Sonnet leads by +3.6
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Claude 3.5 Sonnet
38.7
o1-preview
35.2
MATH level 5
o1-preview leads by +30.0
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Claude 3.5 Sonnet
51.7
o1-preview
81.7
Metr Time Horizons
o1-preview leads by +5.0
Claude 3.5 Sonnet
40.1
o1-preview
45.1
OTIS Mock AIME 2024-2025
o1-preview leads by +24.6
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Claude 3.5 Sonnet
6.4
o1-preview
31.0
SimpleBench
o1-preview leads by +17.0
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Claude 3.5 Sonnet
13.0
o1-preview
30.0
WeirdML
o1-preview leads by +16.6
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Claude 3.5 Sonnet
31.0
o1-preview
47.6
Full benchmark table
| Benchmark | Claude 3.5 Sonnet | o1-preview |
|---|---|---|
Aider · Code Editing | 84.2 | 79.7 |
Chatbot Arena Elo · Overall | 1374.0 | 1388.5 |
Cybench Cybench · evaluates AI on real Capture-The-Flag cybersecurity challenges, testing vulnerability analysis, exploitation, and security reasoning. | 17.5 | 10.0 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 38.7 | 35.2 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 51.7 | 81.7 |
Metr Time Horizons | 40.1 | 45.1 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 6.4 | 31.0 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 13.0 | 30.0 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 31.0 | 47.6 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| n/a | n/a | n/a | n/a | |
| n/a | n/a | n/a | n/a |