Preview build — figures marked “✓” or in green are sourced (benchmarks by Artificial Analysis); treat everything else (value scores, “Est.” figures, unmarked prices and changes) as illustrative. Methodology
Pricing intelligence
Model Price Book
The price book for AI models — sourced prices for LLMs, video, TTS & image, with provenance on every number.
- Models
- 31
- Providers
- 12
- Categories
- 4
- Recent changes
- 3
Four markets
Every modality prices differently.
Text and reasoning APIs priced per million tokens.
Cheapest (sourced)
GPT-4.1 nano
$0.10/1M in ✓
Text-to-video and image-to-video, priced per second of output.
Cheapest (sourced)
Seedance 1.5
$0.052/sec ✓
Speech synthesis priced per thousand characters of input.
Cheapest (sourced)
Cloud TTS Standard
$0.0040/1K ✓
Text-to-image models priced per generated image.
Cheapest (sourced)
FLUX Schnell
$0.003/img ✓
What changed
Recent price moves.
Scheduled: Anthropic's published introductory pricing for Claude Sonnet 5 ends Aug 31, 2026 — input rises $2 → $3 and output $10 → $15 per 1M tokens from Sep 1, per Anthropic's pricing docs.✓ source
Illustrative sample: output price trimmed from $16 to $14 per 1M tokens.
Illustrative sample: output price lowered from $3.00 to $2.50 per 1M tokens.
Efficient frontier
Nothing cheaper and better.
Each category's frontier: the models no tracked rival beats on both price and quality at once. No composite score, no normalisation — pick your quality bar and take the frontier model at that level. Quality data by Artificial Analysis; models without a sourced score are excluded.
- Gemini 2.5 Flash-LiteGoogle$0.10/1M in
- Qwen 3 235BTogether$0.20/1M in
- Gemini 2.5 ProGoogle$1.25/1M in
- o3OpenAI$2.00/1M in
- Claude Sonnet 5Anthropic$2.00/1M in
- Seedance 1.5fal.ai$0.052/sec
- Kling 2.5fal.ai$0.070/sec
- Veo 3.1 FastGoogle$0.100/sec
- Kling 3.0Replicate$0.168/sec