Curiosity · AI Model
OpenAI o1 Pro
o1 Pro Mode shipped with ChatGPT Pro in December 2024 as the premium reasoning tier above o1. It allocates substantially more thinking tokens per response, producing answers that beat o1 on the hardest math, science, and code evaluations — at the cost of latency measured in tens of seconds per reply.
Model specs
- Vendor
- OpenAI
- Family
- o1
- Released
- 2024-12
- Context window
- 128,000 tokens
- Modalities
- text, vision
Strengths
- Highest answer quality in the o1 family at launch
- Allocates extra inference thinking tokens automatically
- Strong on adversarial math and chain-of-thought benchmarks
Limitations
- Latency can exceed 30–60 seconds per answer
- Only available via ChatGPT Pro tier, not the standard API
- Superseded for most workflows by o3, o3-mini and GPT-5 Thinking
Use cases
- Hard-math tutoring and olympiad-style problem solving
- Deep scientific-paper reasoning and synthesis
- Complex agent planning where latency is acceptable
- Engineering design review on critical code
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| AIME 2024 | ≈80%+ | 2024-12 |
| GPQA Diamond | ≈78% | 2024-12 |
| Codeforces rating | top-tier (≈1900) | 2024-12 |
Frequently asked questions
What is o1 Pro?
o1 Pro (sometimes called o1 Pro Mode) is the highest-quality tier of OpenAI's o1 reasoning series. It uses additional inference compute per response to answer hard math, science, and coding problems more reliably than base o1.
How do I access o1 Pro?
o1 Pro Mode ships inside ChatGPT Pro — OpenAI's USD 200/month subscription tier — with limited exposure through the API. Enterprise customers can often request access through sales.
Is o1 Pro still the best reasoning model?
No — o3, o3-mini, and GPT-5 Thinking surpass o1 Pro on most 2025-era reasoning evals. o1 Pro remains a useful baseline but is no longer state-of-the-art.
Sources
- OpenAI — o1 System Card — accessed 2026-04-20
- OpenAI — ChatGPT Pro announcement — accessed 2026-04-20