Buy AI inference tokens at spot, or lock the price ahead with monthly forwards. Open 24/7.
| Model | Input / M | Output / M | Latency | Uptime |
|---|---|---|---|---|
| DeepSeek V4 Flash | $0.20 | $0.40 | ~918ms | 100% |
| Qwen3.8 Max | $2.00 | $6.00 | ~50ms | 100% |
| Kimi K2.7 Code | $0.95 | $4.00 | ~67ms | 100% |
| GLM-5.2 | $1.40 | $4.40 | ~576ms | 100% |
| Kimi K3 | $3.00 | $15.00 | ~456ms | 100% |
GLM-5.2 · 20260731 · daily
Delivery Period: Aug 01, 2026 - Aug 31, 2026
OST · Output-Standardized Tokens
GPU hours are the input; tokens are the output. The gap between them is where AI margins live. Both indices update daily from live rate cards — published as benchmarks, not yet tradeable.
TPI · Token Price Index
87 modelsBlended $/1M tokens across frontier models
30D Daily Updated
GPI · GPU Price Index
40 providersOn-demand $/GPU/hr across 30+ providers
30D Daily Updated
The company that publishes a rate card is the company on the hook when a forward expires — quoting and delivering sit with the same entity. No scraped estimates, no stale quotes.
Listed providers
