Ranked by output cost per million tokens. The premium tier of AI pricing — what the most powerful models actually cost.
Models are ranked by output cost per million tokens (highest first). Output price is the primary cost driver for most AI workloads (chat, code generation, content). Only active, non-deprecated models are included. All prices sourced directly from provider pricing pages.
OpenAI's flagship reasoning model. The most expensive production API in the industry — designed for tasks where accuracy matters more than cost.
OpenAI's previous-gen Pro model. Same pricing as 5.5 Pro — the Pro tier carries a massive premium regardless of generation.
Anthropic's creative specialist model. 1M context window with a steep output premium for creative and narrative tasks.
Anthropic's invitation-only model via Project Glasswing. Not generally available — limited access for select partners.
OpenAI's latest standard model. The new generation baseline — still carries a significant output premium.
OpenAI's previous flagship. Same pricing as GPT-5.6 Sol — the standard tier pricing has remained stable across generations.
Anthropic's most capable model. The Opus tier commands the highest per-token price in Anthropic's lineup.
Anthropic's previous-gen Opus. Same pricing tier as 4.8 — Anthropic keeps Opus pricing consistent across generations.
OpenAI's mid-tier model. A step down from Sol but still commands a significant output premium over budget models.
OpenAI's standard previous-gen model. Same pricing tier as GPT-5.6 Terra — the mid-range has held steady.
The most expensive model costs 1,290x more per output token than the cheapest. Here's the real dollar difference at 1M output tokens/month.
At 100M output tokens/month: $18,000 vs $14. That's a $17,986/mo difference.
Use our free calculator to compare all 67 models and find the sweet spot between capability and cost.