Model Details
o3-mini (high)
Quick answer: o3-mini at high reasoning effort is OpenAI's January 2025 cost-efficient reasoning model configured for maximum accuracy. It scores 67.4% on LiveCodeBench — the highest result among models in its price range at launch. Priced at $1.10/$4.40 per 1M tokens, it was the predecessor to o4-mini and remains available for teams requiring a specific versioned o3-mini deployment.
At a Glance
Where o3-mini (high) leads
- 67.4% LiveCodeBench — strong competitive programming at low cost
- $1.10/$4.40 per 1M tokens — same price as o4-mini, with high-effort reasoning
- Fast iteration cycle: smaller model = lower latency than o3
- Text-only: clean, predictable outputs without multimodal complexity
Where it lags
- Superseded by o4-mini, which achieves 74.2% LiveCodeBench at the same price
- Text-only: no vision capability
- No multimodal inputs
- January 2025 knowledge cutoff predates many 2025 events
Best for: Existing o3-mini integrations; historical benchmark comparisons; teams with validated production workflows on the specific o3-mini model version.
What o3-mini (high) Is
o3-mini is OpenAI's January 2025 compact reasoning model — a predecessor to o4-mini, sharing the chain-of-thought reasoning architecture but trained on an earlier data mix and at smaller scale. The "high" effort setting allocates the maximum reasoning token budget, trading latency for accuracy.
At launch, o3-mini-high set new records for cost-efficient reasoning: its 67.4% LiveCodeBench score outperformed much larger and more expensive models at the time. It was the standard choice for teams needing o-series reasoning without o3's premium price.
o4-mini replaced o3-mini as the recommended cost-efficient reasoning model in April 2025, improving LiveCodeBench from 67.4% to 74.2% at the same price point and adding multimodal vision support.
Specifications
| Field | Value |
|---|---|
| Organization | OpenAI |
| Parameters | Undisclosed |
| Context window | 200,000 tokens |
| Effort setting | High (maximum reasoning budget) |
| Architecture | Chain-of-thought reasoning (o-series) |
| License | Proprietary (API only) |
| Release date | January 31, 2025 |
| Knowledge cutoff | June 2024 (estimated) |
| Modality | Text only |
Pricing
| Input (per 1M tokens) | Output (per 1M tokens) | |
|---|---|---|
| OpenAI API | $1.10 | $4.40 |
Reasoning tokens billed as output. Pricing per OpenAI pricing page.
Public Benchmark Scores
| Benchmark | Score | Source | Date |
|---|---|---|---|
| LiveCodeBench | 67.4% (high effort) | Benchgen evaluation | 2025-01 |
o3-mini (high) vs Alternatives
| Model | LiveCodeBench | Vision | Price (in/out per 1M) |
|---|---|---|---|
| o3-mini (high) | 67.4% | No | $1.10 / $4.40 |
| o4-mini | 74.2% (medium) | Yes | $1.10 / $4.40 |
| o4-mini (high) | 80.2% (high) | Yes | $1.10 / $4.40 |
o4-mini strictly dominates o3-mini-high at the same price: higher LiveCodeBench, vision support, newer knowledge cutoff. For any new deployment, o4-mini is the recommended choice.
Related Models & Benchmarks
- Compare: o4-mini · o4-mini (high) · o3
- Benchmarks: LiveCodeBench
Frequently Asked Questions
What is o3-mini (high)?
o3-mini at high effort is OpenAI's January 2025 compact reasoning model at maximum accuracy setting, scoring 67.4% on LiveCodeBench. It has been superseded by o4-mini, which scores 74.2% at the same price.What is the difference between o3-mini high and o4-mini?
o4-mini scores 74.2% on LiveCodeBench vs o3-mini-high's 67.4%, supports vision input, and has a newer knowledge cutoff — all at the same $1.10/$4.40 per 1M price. o4-mini is strictly better for new deployments.Is o3-mini still available?
Yes. o3-mini remains available via the OpenAI API for teams with existing integrations. For new deployments, o4-mini is the recommended replacement.Specs and scores sourced from OpenAI's official o3-mini announcement (January 2025) and Benchgen evaluations. Pricing cited to the OpenAI pricing page. Last updated 2026-07-24.
Benchmark Leaderboards
This model isn’t on any benchmark leaderboard yet.