| Model | Model ID | Input / 1M | Cached / 1M | Output / 1M | Basis |
|---|---|---|---|---|---|
| GLM-5.3 | glm-5.3 | $0.15 | — | $0.52 | Reliova public rate; cached input is billed as standard input unless stated otherwise. |
| GLM-5.2 | glm-5.2 | $0.15 | — | $0.52 | Reliova public rate; cached input is billed as standard input unless stated otherwise. |
| Kimi-K3 | kimi-k3 | $2.98 | $0.30 | $14.90 | Reliova public rate for the listed token categories. |
| DeepSeek V4 Pro | deepseek-v4-pro | $0.60 | $0.022 | $2.01 | Reliova public rate; thinking configuration can materially change billed output. |
| DeepSeek V4 Flash | deepseek-v4-flash | $0.22 | $0.0070 | $0.67 | Reliova public rate; optimized for cost-sensitive, latency-sensitive traffic. |
Reliova published rate. Updated Sep 15, 2026 · 18:43 UTC. This page requests Reliova’s live price API each time it loads. Rates can change and should be confirmed before funding.
How to compare responsibly
Use the table as an input—not as the decision. A model that costs less per token may produce more reasoning tokens, require retries, miss your latency target, or fail enough cases to cost more overall. Reliova compares candidates on a fixed workload and reports price and performance separately.
How purchasing works
Choose a model, request the current selling rate, and fund a small prepaid balance. Usage is deducted by billed token category. After your application confirms quality and performance, top up the balance or ask for larger-volume pricing.