Cross-provider insight

Where the market is cheap — and where it bites

Every provider's median price, cheapest to priciest — then the models where the output token bill quietly dwarfs the input. The kind of comparison you can only make when every provider sits on one schema.

Provider price ladder

Median input price across each provider's active lineup, per million tokens. Cheapest on top.

1 Mistral AI
$0.15/M in
2 AWS Bedrock
$0.43/M in$1.72/M out
3 Google Gemini
$0.5/M in$3/M out
4 OpenAI
$1.18/M in$8/M out
5 xAI
$1.25/M in$2.5/M out
6 Cohere
$2.5/M in$10/M out
7 Anthropic
$5/M in$25/M out
Bar length is the median input price on a log scale. Median, not mean — one outlier SKU won't drag a whole provider up or down.

⚠️ The output trap

You budget on input tokens — then the output bill lands. These models charge the steepest premium to generate versus to read.

20×
Gemini 3.1 Flash TTS Preview
Google Gemini
in $1 → out $20
8.3×
Gemini 2.5 Flash
Google Gemini
in $0.3 → out $2.5
8×
GPT-5
OpenAI
in $1.25 → out $10
8×
GPT-5 Mini
OpenAI
in $0.25 → out $2
8×
GPT-5 Nano
OpenAI
in $0.05 → out $0.4
8×
GPT-5 Pro
OpenAI
in $15 → out $120
Coming into focus

This ranks the market as it stands today. We re-verify these prices continuously — so the moment a provider moves, this page gains a "who's cutting fastest" view: the biggest movers, ranked, with the change since we started tracking. Price history you can't buy after the fact.

How this is built

Per-provider figures are the median of verified, published per-million-token prices across that provider's active models. The output trap is simply output ÷ input for each model — no weighting, no score. Retired models are excluded. Every value links to the provider's own pricing page.

Updated Jun 3, 2026 18:32 UTC · Full methodology →