o3

By price, o3 ranks 48th cheapest of 68 on input and 49th on output. Its quality score of 95 ranks 6th of 68, and on value (quality score ÷ output price) it ranks 45th of 68. Its blended rate sits 19% above the catalog average.

In real billing terms: a support chatbot handling 50,000 turns a month (2,000 input / 500 output tokens each) costs about $400 per month on o3, while a RAG workload of 20,000 queries (8,000 input / 800 output tokens each) runs about $448. Because output is priced 4.0× input here, long-generation tasks scale cost fastest — cap output length before you optimise anything else.

A 200K context window holds roughly 267 pages of A4 text (at ~750 tokens per page), ranking 44th of 68 here. That covers typical RAG and multi-turn chat, though book-length inputs still need retrieval and chunking. Output runs at about 40 tok/s (64th fastest of 68), which is better suited to asynchronous or background jobs.

We track 11 OpenAI models, and o3's blended rate sits 8% below that provider's average. If your task tolerates a little less headroom, deepseek-reasoner-v4 (DeepSeek) blends about 84% cheaper at a quality score of 92, and is worth benchmarking first. Going the other way, gpt-5.6-sol scores 98 on quality for about 129% more.

The closest model in OpenAI's line-up is gpt-4o ($2.50 input / $10.00 output, quality 88). Against it, o3 scores 7 points higher on quality, runs 60 tok/s slower (about 60%), has a 72K larger context window, costs 25% less blended. When two models sit this close on price, the deciding factor is usually the quality ceiling — run both against your real prompts before picking.