OpenAI O1 Mini Preview: Verified Release Facts
OpenAI released o1-mini and o1-preview on September 12, 2024. These models use reinforcement learning for complex reasoning via the v1/chat/completions API.
Release Announcement and Scope
The official announcement date for this model release is September 12, 2024. This historical date distinguishes the event from the current publication date of this article. On that day, OpenAI documented the release of two new large language models: o1-preview and o1-mini. The source identifies these models as part of the OpenAI o-series family. The primary stated purpose of these models is to perform complex reasoning tasks using reinforcement learning training methods.
- Announcement Date: September 12, 2024
- Models Released: o1-preview and o1-mini
- Model Family: OpenAI o-series
- Training Method: Reinforcement Learning
Confirmed Technical Specifications
The verified facts establish specific technical parameters for o1-mini. The model accepts both text and image inputs while producing text outputs. Access to the model is provided through the v1/chat/completions API endpoint. The source records that open weights are not confirmed for this model. This status indicates a lack of confirmation regarding weight availability; it does not establish that local installation is impossible, nor does it confirm that API access is the only method of interaction. No pricing information, context window size, or maximum output token limit is present in the source facts.
- Input Modalities: Text and Image
- Output Modality: Text
- API Endpoint: v1/chat/completions
- Open Weights Status: Not Confirmed
Proposed Evaluation Procedure
Engineers can evaluate o1-mini by designing a controlled test suite focused on its stated capability for complex reasoning. First, select a set of multi-step logic problems to test the model's reasoning performance. Second, measure the accuracy of the responses by comparing them against known correct solutions. Third, validate the image input capability by processing varied visual datasets. Finally, compare the results against baseline models to observe distinct behaviors. This procedure allows for empirical assessment without assuming benefits not explicitly stated in the source documentation.
- Test: Execute complex reasoning prompts via the API
- Measure: Calculate accuracy rates against ground truth data
- Validate: Confirm image input processing capabilities
- Compare: Analyze results against existing baselines
Artificial Analysis benchmark review
Benchmark review updated on . These scores apply to the linked artifact and recorded methodology; they do not establish the performance of a different runtime or quantized model.
| Measurement | Recorded result | Measured |
|---|---|---|
| Artificial Analysis Intelligence Index (v4.1) ↗ | 14 index_points | 2026-08-09 |
| Artificial Analysis Intelligence Index (v4.2) ↗ | 8.1 index_points | 2026-09-05 |
| Artificial Analysis Intelligence Index (v4.3) ↗ | 9.8 index_points | 2026-09-09 |
Review the full benchmark provenance and source methodology →
Primary sources
Frequently asked questions
What is the stated training method for o1-mini?
The source states that o1-mini is trained with reinforcement learning to perform complex reasoning tasks.
Does the source confirm local installation options?
No. The source only records that open weights are not confirmed. It does not state whether local installation is possible or impossible.