On the cheapest model from each, Anthropic wins on raw output price: Claude Haiku 3.5 at $4 per 1M output tokens versus $4.5. But "cheaper per token" rarely means "cheaper for your app". The model that needs fewer retries, shorter prompts, or less reasoning can cost less in practice even at a higher sticker rate.
Last verified June 14, 2026| Model | In / 1M | Out / 1M |
|---|---|---|
| GPT-5.5 | $5 | $30 |
| GPT-5.4 | $2.5 | $15 |
| GPT-5.4 mini | $0.75 | $4.5 |
| GPT-Realtime-2 audio | $32 | $64 |
| GPT-Realtime-2 text | $4 | $24 |
| GPT-Realtime-2 image | $5 | — |
| GPT-Image-2 image | $8 | $30 |
| GPT-Image-2 text | $5 | — |
| Model | In / 1M | Out / 1M |
|---|---|---|
| Claude Mythos 5 | $10 | $50 |
| Claude Fable 5 | $10 | $50 |
| Claude Opus 4.8 | $5 | $25 |
| Claude Haiku 4.5 | $1 | $5 |
| Claude Haiku 3.5 | $0.8 | $4 |
| Claude Haiku 3 | $0.5 | $6.25 |
| Claude Sonnet 4.5 | $3 | $15 |
| Claude Sonnet 4.6 | $3 | $15 |
| Claude Sonnet 4 | $3 | $15 |
| Claude Opus 4.5 | $5 | $25 |
| Claude Opus 4.6 | $5 | $25 |
| Claude Opus 4.7 | $5 | $25 |
| Claude Opus 4.1 | $15 | $75 |
| Claude Opus 4 | $15 | $75 |
Per-token rates can't answer that. The winner depends on how your code uses each model. PrePrice scans your project and computes the real per-user cost either way, plus what to charge so you clear margin.
Find your real cost — free