OpenAI has introduced o4-mini, a new small-scale reasoning model optimized for speed and affordability while supporting tools like Python, browsing, and image inputs. It features a 200,000-token context window and offers approximately 10x cost reduction compared to the o3 model.
- Pricing is set at $1.10 per million input tokens and $4.40 per million output tokens.
- A configuration called o4-mini-high increases inference effort for higher quality on complex tasks.
- Benchmarks show o4-mini outperforming o3 on AIME 2024 (93.4% vs 91.6%) and Codeforces (ELO 2719 vs 2706).
- It trails o3 slightly on SWE-Bench Verified (68.1% vs 69.1%) and MMMU (81.6% vs 82.9%).
The model provides comparable performance to larger models like o3 at a significantly lower cost, making it suitable for tool-augmented reasoning tasks.