Compare model prices
Input, output, cached and batch prices, every route and a month of tokens, side by side.
Qwen2.5 72B InstructQwen3.8 Max PrimeA month of tokens
100M input and 20M output tokens, list price. Use your own numbers.
Side by side
| Qwen2.5 72B Instruct | Qwen3.8 Max Prime | |
|---|---|---|
| Input per 1M | $0.36 | $4 |
| Output per 1M | $0.4 | $12 |
| Cached input | n/a | $0.5 |
| Batch in / out | n/a | n/a |
| Blended (3:1) | $0.37 | $6 |
| Context | 33K | 1M |
| Routes | 2 | 1 |
| Released | Sep 19, 2024 | Sep 23, 2026 |
| Leadersboard score | n/a | n/a |
On the price map
Size: context window, 8K to 2MEqual blended cost (3:1) 243 models, lower left is cheaper
Every route
Qwen2.5 72B Instruct
| Way to buy | Input | Output | In / out per 1M | Cached | Vs list |
|---|---|---|---|---|---|
 DeepInfrafp8 | $0.36 | $0.4 | $0.36 / $0.4Same as list | n/a | Same as list |
 Novita AIbf16 | $0.38 | $0.4 | $0.38 / $0.4+4% vs list | n/a | +4% vs list |
Qwen3.8 Max Prime
| Way to buy | Input | Output | In / out per 1M | Cached | Vs list |
|---|---|---|---|---|---|
| Qwen (Alibaba) API The lab's own API · cached $0.5 | $4 | $12 | $4 / $12List price | $0.5 | List price |
List prices from providers' pages on the date shown; your contract may differ. Logos via logo.dev; trademarks belong to their owners.