Agent Metering and Billing
This topic describes the various billing items and related costs of the AI agent.
Credits
The platform uses credits as the unified billing unit. Users must first purchase credits through value-added services, and the corresponding amount will be deducted based on actual usage of billable items.
Billing cycle
All billable items are charged on a daily basis. The platform calculates the credits consumed each day according to actual usage. Invoices are typically generated two days after the end of the current billing cycle, with the exact timing subject to system processing.
Fee structure
AI agents are charged on a pay-as-you-go basis according to the following formula:
Total fee = Model fee + AI voice fee + Extended capability fee - Waived fee
For details about each billing item, see the following sections.
Model fees
Available models
For more information about how to differentiate models, see Available Models.
Billing formula
Model services are charged based on the model’s token usage. The billing formula is as follows:
Model fees = Token usage × Unit price of tokens
In a large language model, a token is the basic unit of text processing. The model usually breaks down the input text into a series of tokens and then processes and analyzes these tokens. Tokens can be words, characters, subword fragments, or other text segments. The specific segmentation is determined by the model’s tokenization algorithm. Therefore, token calculation and processing methods might vary depending on the model’s architecture and design.
Unit prices
Qwen
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| Qwen-max input | 2.50 | ¥2.50 |
| Qwen-max output | 10.00 | ¥10.00 |
| Qwen-turbo-latest input | 0.36 | ¥0.36 |
| Qwen-turbo-latest output | 1.40 | ¥1.40 |
| Qwen3-32b input | 0.75 | ¥0.75 |
| Qwen3-32b output | 7.50 | ¥7.50 |
| Qwen3-max input | 2.50 | ¥2.50 |
| Qwen3-max output | 10.00 | ¥10.00 |
| Qwen-flash input | 0.15 | ¥0.15 |
| Qwen-flash output | 1.50 | ¥1.50 |
| Qwen-plus input | 0.80 | ¥0.80 |
| Qwen-plus output | 2.00 | ¥2.00 |
Doubao
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| Doubao-seed-1.6-flash input | 0.30 | ¥0.30 |
| Doubao-seed-1.6-flash output | 3.00 | ¥3.00 |
| Doubao-seed-1.6 input | 0.80 | ¥0.80 |
| Doubao-seed-1.6 output | 2.00 | ¥2.00 |
| Doubao-seed-1.8 input | 0.80 | ¥0.80 |
| Doubao-seed-1.8 output | 2.00 | ¥2.00 |
| Doubao-seed-2.0-mini input | 0.20 | ¥0.20 |
| Doubao-seed-2.0-mini output | 2.00 | ¥2.00 |
| Doubao-seed-2.0-Pro input | 3.20 | ¥3.20 |
| Doubao-seed-2.0-Pro output | 16.00 | ¥16.00 |
DeepSeek
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| DeepSeek v3 input | 0.20 | ¥0.20 |
| DeepSeek v3 output | 3.00 | ¥3.00 |
MiniMax
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| MiniMax-m2.7 input | 2.20 | ¥2.20 |
| MiniMax-m2.7 output | 8.70 | ¥8.70 |
ChatGPT
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| GPT-4o input | 16.67 | $2.50 |
| GPT-4o output | 66.67 | $10.00 |
| GPT-4o-mini input | 1.00 | $0.15 |
| GPT-4o-mini output | 4.00 | $0.60 |
| GPT-5 input | 8.34 | $1.25 |
| GPT-5 output | 66.67 | $10.00 |
| GPT-5-mini input | 1.67 | $0.25 |
| GPT-5-mini output | 13.34 | $2.00 |
| GPT-5-nano input | 0.33 | $0.05 |
| GPT-5-nano output | 2.67 | $0.40 |
| GPT-5.1 input | 8.34 | $1.25 |
| GPT-5.1 output | 66.67 | $10.00 |
| GPT-5.2 input | 11.67 | $1.75 |
| GPT-5.2 output | 93.40 | $14.00 |
| GPT-5.4 input | 33.35 | $5.00 |
| GPT-5.4 output | 200.00 | $30.00 |
| GPT-5.4-mini input | 5.00 | $0.75 |
| GPT-5.4-mini output | 30.00 | $4.50 |
| GPT-5.4-nano input | 1.33 | $0.20 |
| GPT-5.4-nano output | 8.34 | $1.25 |
Gemini
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| Gemini-2.0-flash input | 0.67 | $0.10 |
| Gemini-2.0-flash output | 2.67 | $0.40 |
| Gemini-2.5-pro input | 8.40 | $1.25 |
| Gemini-2.5-pro output | 66.67 | $10.00 |
| Gemini-2.5-flash input | 2.00 | $0.30 |
| Gemini-2.5-flash output | 16.65 | $2.50 |
| Gemini-3-flash input | 3.34 | $0.50 |
| Gemini-3-flash output | 20.00 | $3.00 |
| Gemini-3.1-pro input | 13.34 | $2.00 |
| Gemini-3.1-pro output | 80.00 | $12.00 |
Mistral
| Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|
| Mistral-large-latest input | 53.36 | $8.00 |
| Mistral-large-latest output | 160 | $24.00 |
AI voice fees
Billing formula
AI voice services consist of speech input (Automatic Speech Recognition, ASR) and speech output (Text-to-Speech, TTS). The billing formulas are as follows:
ASR fee = ASR unit price × Input audio duration
TTS fee = TTS unit price × Output character count
ASR: A technology that recognizes and understands natural human speech input. By analyzing and processing audio signals, it converts speech into text.
TTS: A technology that converts text into spoken audio output. It simulates human speech to transform written text into audio information.
Unit prices
| ASR provider | ASR model | Credits per hour | Amount per hour |
|---|---|---|---|
| ALIYUN | paraformer-realtime-v2 | 0.13 | ¥0.13 |
| TENCENT | 16k_zh_en | 3.20 | ¥3.20 |
| VOLCANO | volcengine_streaming_common | 3.50 | ¥3.50 |
| VOLCANO | bigmodel | 4.50 | ¥4.50 |
| AZURE | azure-stt-standard | 6.67 | $1.00 |
| ELEVENLABS | scribe_v1_experimental | 1.47 | $0.22 |
| TTS provider | TTS model | Credits per 1,000 characters | Amount per 1,000 characters |
|---|---|---|---|
| ALIYUN | cosyvoice-v3-plus | 2.00 | ¥2.00 |
| ALIYUN | cosyvoice-v3-flash | 1.00 | ¥1.00 |
| TENCENT | default | 9.00 | ¥9.00 |
| VOLCANO | seed-tts-1.0 | 5.00 | ¥5.00 |
| VOLCANO | seed-tts-2.0 | 3.00 | ¥3.00 |
| AZURE | neural | 100.00 | $15.00 |
| AZURE | multilingual-neural | 100.00 | $15.00 |
| AZURE | dragon-hd-latest | 146.67 | $22.00 |
| AZURE | dragon-hd-flash | 146.67 | $22.00 |
| chirp3-hd | 200.00 | $30.00 | |
| studio | 1066.72 | $160.00 | |
| neural2 | 106.67 | $16.00 | |
| wavenet | 26.67 | $4.00 | |
| polyglot | 106.67 | $16.00 | |
| standard | 26.67 | $4.00 | |
| MINIMAX | speech-02-turbo | 2.00 | ¥2.00 |
Extended capabilities fees
Depending on your agent configuration, in addition to the basic AI resource consumption, your product may require other extended AI capabilities.
Configuration:
- My Agent > Develop > Model Configuration
- Workflow Management
Timbre cloning TTS fees
As shown in the image, after enabling and publishing the Timbre Cloning feature for the agent, users will be able to interact via voice using the cloned voice. Subsequent fees will be incurred based on the actual usage of this conversation.
| Vendor | Model | Credits per 10,000 characters | Amount per 10,000 characters |
|---|---|---|---|
| VOLCANO | seed-icl-1.0 | 8.00 | ¥8.00 |
| ALIYUN | cosyvoice-v3-plus | 2.00 | ¥2.00 |
| ALIYUN | cosyvoice-v3-flash | 1.00 | ¥1.00 |
| AZURE | DragonLatestNeural | 146.67 | $22.00 |
| google-voice-clone | 400.00 | $60.00 |
AI image generation fees
As shown in the image, after configuring and publishing the image generation node for the workflow, the agent will be empowered with the ability to generate images. Subsequent fees will be incurred based on the actual usage of this image generation operation.
This node will be available soon. Stay tuned.
| Vendor | Model | Credits per image | Amount per image |
|---|---|---|---|
| VOLCANO | doubao-seedream-4.0 | 0.20 | ¥0.20 |
| VOLCANO | doubao-seedream-5.0-lite | 0.22 | ¥0.22 |
| ALIYUN | z-image-turbo | 0.10 | ¥0.10 |
| gemini-2.5-flash-image | 0.27 | $0.04 |
Web search fees
As shown in the image, after configuring and publishing the Web Search tool for the agent or workflow, it will be empowered with the ability to retrieve real-time web information. Subsequent fees will be incurred based on the actual usage of this search operation (such as querying daily news).
| Vendor | Model | Credits per 1,000 requests | Amount per 1,000 requests |
|---|---|---|---|
| VOLCANO | Volcano Colab (Pay-as-you-go) | 30.00 | ¥30.00 |
| BRAVE | brave | 53.34 | $8.00 |
Historical conversation summary fees
As shown in the image, after enabling and publishing the Historical conversation summary feature for the agent, it will have the ability to analyze historical conversations. Subsequent fees for this capability will be incurred based on usage.
| Vendor | Model | Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|---|---|
| ALIYUN | qwen-plus | Input | 0.80 | ¥0.80 |
| ALIYUN | qwen-plus | Output | 2.00 | ¥2.00 |
| gemini-2.5-pro | Input | 8.33 | $1.25 | |
| gemini-2.5-pro | Output | 66.67 | $10.00 |
Event memory fees
As shown in the image, after enabling and publishing the Conversation Event Memory feature for the agent, the agent can remember event history over the long term. This provides a continuous conversational experience where the agent remembers what happened a long time ago. Subsequent fees for this feature will be incurred based on usage.
| Vendor | Model | Billing item | Credits per million tokens | Amount per million tokens |
|---|---|---|---|---|
| ALIYUN | qwen3-max | Input | 2.50 | ¥2.50 |
| ALIYUN | qwen3-max | Output | 10.00 | ¥10.00 |
| OPENAI | gpt-5.1 | Input | 8.33 | $1.25 |
| OPENAI | gpt-5.1 | Output | 66.67 | $10.00 |
Fee waivers
If the agent is deployed to a device for direct connection, the following fee waivers apply:
Basic AI fee waiver
If you have enabled the AI Agent Integration advanced feature while developing the product, a certain quota will be waived when the device incurs daily fee consumption. The waiver quota is 0.5 resource credits per day. Once the waiver quota is exhausted, any excess daily consumption will be billed.
Subscription model fee waiver
If you enroll the corresponding product in the Subscription Model, Tuya will provide unified device-side subscription plans, benefit distribution, usage statistics, and more. In addition to the AI basic fee waiver, your product is eligible for further fee waivers, potentially achieving a full fee waiver.
Is this page helpful?
YesFeedbackIs this page helpful?
YesFeedback
