TL;DR — GPT-4o costs ~17× GPT-4o-mini ($2.50/$10 vs $0.15/$0.60) — this is OpenAI's biggest same-family gap. Every number below is from our weekly-verified pricing tracker (verified 2026-08-26).
Pricing side by side (per 1M tokens, USD)
| Model | Provider | Input | Output | |-------|----------|-------|--------| | GPT-4o | OpenAI | $2.5 | $10 | | GPT-4o-mini | OpenAI | $0.15 | $0.6 |
Verified 2026-08-26 against official rate cards — the tracker keeps the change log.
What a real month costs
At a standard workload of 10M input + 2M output tokens/month:
- GPT-4o: $45.00
- GPT-4o-mini: $2.70
On this mix, GPT-4o-mini is the cheaper column (16.7× gap). Your ratio will differ — run your own mix through the LLM cost calculator.
Choose GPT-4o if
quality and multimodal depth matter for user-facing output; 4o holds the line where mini's ceiling shows.
Choose GPT-4o-mini if
the job is high-volume and mechanical: classification, tagging, extraction, simple summaries. At $0.15/$0.60, mini is the reason 'just run everything through the big model' is usually a budget mistake.
Routing decision card
- The 17× spread is the single strongest argument for routing inside OpenAI's own lineup
- Mini absorbs bulk; 4o (or above) takes user-visible output — misrouting even 20% of bulk traffic up-tier multiplies that slice's cost 17×
- Mini details and edge cases: see the GPT-4o-mini deep dive
- The router answer: most teams shouldn't hard-pick one — ClawRouters routes each request to whichever side fits it, so the comparison becomes a per-call decision instead of a bet.
Per-model detail: GPT-4o-mini deep dive
Part of the LLM model comparison index. Prices move fast in 2026 — the tracker logs every change with dates.