One USD invoice
When an application uses DeepSeek for routine work and Kimi or Qwen for a specialist path, the team funds one balance instead of reconciling a separate payment account for each provider.
Search or filter the current AIWave route catalog. The table keeps input, cache-hit input, and output separate so one blended number cannot hide the expensive part of a workload.
Checked 2026-09-10. USD per 1M text tokens. Listed rates use default ×1; keys created with the VIP billing group use ×0.9.
Filtered routes are shown below.
| Model ID | Provider | Input | Cache hit | Output | Effective |
|---|---|---|---|---|---|
deepseek-flash |
DeepSeek | $0.30 | $0.006 | $1.20 | 2026-09-10 |
deepseek-v3.2 |
DeepSeek | $0.154 | Not listed | $0.308 | 2026-08-27 |
deepseek-v4-flash |
DeepSeek | $0.30 | $0.006 | $1.20 | 2026-08-27 |
deepseek-v4-pro |
DeepSeek | $1.32 | $0.044 | $3.96 | 2026-08-27 |
glm-4.5 |
GLM | $0.6975 | $0.18 | $1.207 | 2026-08-27 |
glm-4.5-air |
GLM | $0.6525 | $0.162 | $2.03 | 2026-08-27 |
glm-4.6 |
GLM | $0.93 | $0.22 | $3.41 | 2026-08-27 |
glm-4.7 |
GLM | $0.93 | $0.22 | $3.41 | 2026-08-27 |
glm-5 |
GLM | $1.55 | $0.400001 | $4.96 | 2026-08-27 |
glm-5-turbo |
GLM | $1.8 | $0.480001 | $5.4 | 2026-08-27 |
glm-5.1 |
GLM | $2.1 | $0.680001 | $6.6 | 2026-08-27 |
kimi-k2.5 |
Kimi | $0.66 | $0.121846 | $3.3 | 2026-08-27 |
kimi-k2.6 |
Kimi | $1.09 | $0.186857 | $4.5998 | 2026-08-27 |
kimi-k2.7-code |
Kimi | $1.89 | $0.285001 | $6 | 2026-08-27 |
kimi-k2.7-code-highspeed |
Kimi | $4.725 | $0.3012502 | $14.999999 | 2026-08-27 |
kimi-k3 |
Kimi | $4.5 | $0.9 | $22.5 | 2026-08-27 |
moonshot-v1-128k |
Kimi | $1.8 | Not listed | $4.5 | 2026-08-27 |
moonshot-v1-128k-vision-preview |
Kimi | $2.3 | Not listed | $5.75 | 2026-08-27 |
moonshot-v1-32k |
Kimi | $0.95 | Not listed | $2.85 | 2026-08-27 |
moonshot-v1-32k-vision-preview |
Kimi | $1.15 | Not listed | $3.45 | 2026-08-27 |
moonshot-v1-8k |
Kimi | $0.3 | Not listed | $2.2 | 2026-08-27 |
moonshot-v1-8k-vision-preview |
Kimi | $0.3 | Not listed | $2.301 | 2026-08-27 |
moonshot-v1-auto |
Kimi | $75 | Not listed | $75 | 2026-08-27 |
ernie-4.5-turbo |
ERNIE | $0.19726 | $0.049315 | $0.745209 | 2026-08-27 |
ernie-5.0 |
ERNIE | $2.465754 | $0.246575 | $9.315125 | 2026-08-27 |
MiniMax-M2 |
MiniMax | $0.46866 | $0.046866 | $1.87464 | 2026-08-27 |
MiniMax-M2.1 |
MiniMax | $0.46866 | $0.046866 | $1.87464 | 2026-08-27 |
MiniMax-M2.1-highspeed |
MiniMax | $0.47 | $0.046866 | $1.87 | 2026-08-27 |
MiniMax-M2.5 |
MiniMax | $0.46866 | $0.046866 | $1.87464 | 2026-08-27 |
MiniMax-M2.5-highspeed |
MiniMax | $0.47 | $0.046866 | $1.87 | 2026-08-27 |
MiniMax-M2.7 |
MiniMax | $0.45304 | $0.090608 | $1.81216 | 2026-08-27 |
MiniMax-M2.7-highspeed |
MiniMax | $0.45 | $0.0906 | $1.81 | 2026-08-27 |
MiniMax-M3 |
MiniMax | $0.90608 | $0.181216 | $3.62432 | 2026-08-27 |
qwen-72b-chat |
Qwen | $4.463422 | Not listed | $4.463422 | 2026-08-27 |
qwen-audio-3.0-realtime-flash |
Qwen | $0.669513 | Not listed | $0.669513 | 2026-08-27 |
qwen-audio-3.0-realtime-plus |
Qwen | $1.115856 | Not listed | $1.115856 | 2026-08-27 |
qwen-image-2.0-pro-2026-04-22 |
Qwen | $0.122744 | Not listed | $0.122744 | 2026-08-27 |
qwen-image-3.0 |
Qwen | $0.040171 | Not listed | $0.040171 | 2026-08-27 |
qwen-image-3.0-pro |
Qwen | $0.055793 | Not listed | $0.055793 | 2026-08-27 |
qwen1.5-72b-chat |
Qwen | $4.463422 | Not listed | $4.463422 | 2026-08-27 |
qwen3-max |
Qwen | $1.562198 | Not listed | $6.248791 | 2026-08-27 |
qwen3.5-122b-a10b |
Qwen | $0.49315 | Not listed | $3.726044 | 2026-08-27 |
qwen3.5-27b |
Qwen | $0.443836 | Not listed | $3.353447 | 2026-08-27 |
qwen3.5-omni-flash |
Qwen | $0.490976 | Not listed | $2.968176 | 2026-08-27 |
qwen3.5-plus |
Qwen | $0.446342 | Not listed | $2.678053 | 2026-08-27 |
qwen3.6-27b |
Qwen | $0.669513 | Not listed | $4.01708 | 2026-08-27 |
qwen3.6-35b-a3b |
Qwen | $0.401708 | Not listed | $2.410248 | 2026-08-27 |
qwen3.6-flash |
Qwen | $0.267805 | Not listed | $1.606832 | 2026-08-27 |
qwen3.7-flash-2026-07-15 |
Qwen | $0.267805 | Not listed | $1.071221 | 2026-08-27 |
qwen3.7-max-2026-05-20 |
Qwen | $2.678053 | Not listed | $8.03416 | 2026-08-27 |
qwen3.7-max-2026-06-08 |
Qwen | $2.678053 | Not listed | $8.03416 | 2026-08-27 |
qwen3.7-plus |
Qwen | $0.446342 | Not listed | $1.785369 | 2026-08-27 |
qwen3.7-text-embedding |
Qwen | $0.111586 | Not listed | $0.111586 | 2026-08-27 |
qwen3.8-2.4t-a95b |
Qwen | $2.678053 | Not listed | $8.03416 | 2026-08-27 |
qwen3.8-27b |
Qwen | $0.669513 | Not listed | $2.678053 | 2026-08-27 |
qwen3.8-max |
Qwen | $2.678053 | Not listed | $8.03416 | 2026-08-27 |
doubao-seed-2-0-lite-260428 |
Doubao | $0.401708 | Not listed | $2.410248 | 2026-08-27 |
doubao-seed-2-0-mini-260428 |
Doubao | $0.178537 | Not listed | $1.785369 | 2026-08-27 |
doubao-seed-2-1-pro-260628 |
Doubao | $1.339027 | Not listed | $6.695133 | 2026-08-27 |
doubao-seed-evolving |
Doubao | $1.339027 | Not listed | $6.695133 | 2026-08-27 |
step-3.5-flash |
StepFun | $0.21 | $0.042 | $0.63 | 2026-08-27 |
step-3.5-flash-2603 |
StepFun | $0.21 | $0.042 | $0.63 | 2026-08-27 |
step-3.7-flash |
StepFun | $0.4 | $0.08 | $2.4 | 2026-08-27 |
xiaomi/mimo-v2.5-pro |
MiMo | $1.562198 | Not listed | $4.686593 | 2026-08-27 |
Four operating differences you can verify before moving a workload.
When an application uses DeepSeek for routine work and Kimi or Qwen for a specialist path, the team funds one balance instead of reconciling a separate payment account for each provider.
Start with https://aiwave.live/v1 and an explicit
model ID. Changing the model does not require a second provider
SDK or another authentication pattern.
Every public rate carries an effective date, and the same snapshot is available as JSON. A budget review can cite the exact input, cache-hit, and output figures it used.
When a long-context job costs more than expected, inspect the model, input, output, and charge for that request instead of trying to explain a single blended account total.
Provider-family names are not request values. Use the exact ID your test and ledger both show.
A missing cache rate means it is not listed in this snapshot. It does not mean cached input costs zero.
Validate one representative request before forecasting a production workload.