Model Details
o1 Pro
Quick answer: o1 Pro is OpenAI's maximum compute configuration of the o1 reasoning model, released November 20, 2024. It scores 8.12% on Humanity's Last Exam. As the predecessor to o3, o1 Pro targets researchers and advanced users who need the highest possible accuracy from the original o1 model. It has since been superseded by o3 and o4-mini for most use cases.
At a Glance
Where o1 Pro leads
- Higher accuracy than o1 and o1-mini — maximum compute reasoning budget
- Established reasoning model with a stable API history
- Useful for reproducibility of research conducted on o1-series models
Where it lags
- 8.12% Humanity's Last Exam — below o3 (20.32%), o4-mini (18.08%), and Gemini 2.5 Pro (21.64%)
- Superseded by o3 and o4-mini-high for both accuracy and cost-efficiency
- Text-only: no vision support
- October 2023 knowledge cutoff — outdated for current events
- Not available via standard API (ChatGPT Pro/API Pro tier required)
Best for: Research reproducibility on o1-series outputs; teams explicitly benchmarking or comparing against o1 Pro baselines.
What o1 Pro Is
o1 Pro is the "maximum compute" variant of OpenAI's o1 family, providing the longest internal chain-of-thought reasoning for any single o1 query. Released November 20, 2024, it was initially available only to ChatGPT Pro subscribers and was later extended to the API at a premium pricing tier.
The model sits in the same architecture family as o1, using extended reasoning tokens to improve accuracy on multi-step logical, scientific, and mathematical problems. Its 8.12% Humanity's Last Exam score represents o1-class performance with maximum reasoning time — higher than base o1 but below the subsequently released o3 (20.32%).
For any new deployment requiring o-series reasoning, o3 or o4-mini-high provides substantially better accuracy. o1 Pro's primary use case today is backward compatibility: teams that have validated production workflows on the specific o1 Pro model and need reproducible outputs.
Specifications
| Field | Value |
|---|---|
| Organization | OpenAI |
| Parameters | Undisclosed |
| Context window | 200,000 tokens |
| Max output | 32,768 tokens |
| Effort setting | Pro (maximum compute) |
| Architecture | Chain-of-thought reasoning (o1-series) |
| License | Proprietary |
| Release date | November 20, 2024 |
| Knowledge cutoff | October 2023 |
| Modality | Text only |
Pricing
o1 Pro was initially restricted to ChatGPT Pro subscribers ($200/month) before API availability. API pricing was not publicly confirmed at this page's last update — check the OpenAI pricing page for current rates.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| Humanity's Last Exam | 8.12% | Benchgen evaluation | 2024-11 |
o1 Pro vs Alternatives
| Model | HLE | Vision | Price (in/out per 1M) |
|---|---|---|---|
| o1 Pro | 8.12% | No | API tier (see pricing) |
| o3 | 20.32% | No | $10 / $40 |
| o4-mini (high) | 18.08% | Yes | $1.10 / $4.40 |
| Gemini 2.5 Pro | 21.64% | Yes | $1.25 / $10 |
| o1 | 7.96% | No | $15 / $60 |
For new deployments requiring top-tier reasoning, o3 or o4-mini-high are strictly better choices: higher HLE, lower cost, newer knowledge, and (for o4-mini) vision support.
Related Models & Benchmarks
- Compare: o1 · o3 · o4-mini
- Benchmarks: Humanity's Last Exam
Frequently Asked Questions
What is o1 Pro?
o1 Pro is OpenAI's maximum compute o1 model, released November 2024, scoring 8.12% on Humanity's Last Exam. It has been superseded by o3 and o4-mini-high for accuracy and cost-efficiency.What is the difference between o1 and o1 Pro?
o1 Pro uses a larger internal reasoning token budget than base o1, improving accuracy on complex problems. o1 Pro scores 8.12% vs o1's 7.96% on HLE. Both have been superseded by o3 (20.32% HLE).Is o1 Pro still available?
Yes, o1 Pro remains available via the OpenAI API at a premium tier. For most use cases, o3 or o4-mini-high provide better performance at lower cost.What is o1 Pro's knowledge cutoff?
o1 Pro's training knowledge cutoff is October 2023.Specs and scores sourced from OpenAI's official o1 Pro announcement (November 2024) and Benchgen evaluations. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.