DTU Dedicated compute deployment

Dedicated compute deployment offers two options: Dedicated Throughput Unit (DTU) and Model Unit (MU), providing dedicated GPU resources for specified models with performance and throughput guarantees. DTU targets newly released models (billed by input/output TPM × duration); MU is used for existing models (billed by model unit count × duration). This document introduces both options' features, billing, usage flow, and supported models.

Overview

Dedicated compute deployment provides dedicated GPU resources and performance guarantees for specified models, supporting both base models and custom model deployment. Bailian offers two dedicated compute options:

• DTU (Dedicated Throughput Unit): Billed by Input/Output TPM, providing more direct throughput guarantees. Fully managed inference service with dedicated underlying GPU resources maintained by the platform.
• Model Unit (MU): Configures compute power based on usage duration and the number of Model Units, with dedicated resources. Supports custom performance metrics and PD disaggregated computing mode. Billing granularity is model unit count × duration.

DTU and MU are differentiated by model applicability: newly released models use DTU (metered by input/output TPM), while existing models continue to use MU (metered by model unit count). Both provide dedicated compute, resource isolation, and fine-tuned model deployment. For new deployments, choose DTU if the model supports it.

Shared capabilities of both options:

• Dedicated GPU resources; the inference environment is physically isolated from other users.
• Supports base models and custom models (Bailian fine-tuned models or user-uploaded models).
• The platform maintains the underlying operations, no need to manage GPUs.
• No proactive RPM/TPM limits; traffic is bound by actual capacity.

For the general model deployment workflow, see Dedicated Deployment Overview and Provisioned Throughput.

Solution Selection

Bailian offers multiple capacity and billing plans for inference calls. DTU suits scenarios requiring dedicated deployment, fine-tuned model deployment, low latency with high concurrency, or data isolation. The plans are compared in the table below.

For the full comparison of all billing methods including Token pay-as-you-go and PTU, see the main table in Dedicated Deployment Overview.

Comparison Dimension	DTU	PTU	Token Pay-as-you-go	PAI/Lingjun
Feature	Dedicated compute + fine-tuned models	Throughput guarantee + low latency	Elastic and zero-threshold	Custom runtime
Resource isolation	Physical isolation	Logical isolation	Shared pool	Physical isolation
Fine-tuned models	Full-parameter + LoRA	Not supported	LoRA	Supported
Custom framework	Not supported	Not supported	Not supported	Supported
Billing mode	TPM quota × duration	TPM quota × duration	Token usage	GPU × duration

Billing Rules

DTU Billing

DTU is billed separately by input TPM and output TPM, using a pre-paid (monthly) model. Each model has fixed input/output baseline TPM (see table below), and you must purchase in integer multiples of the baseline TPM. Purchase at least 1× baseline TPM for both input and output; for production, 2× or above is recommended for each.

Fees are calculated by the backend based on model, input/output TPM, service region, and purchase duration. Prices for each model are shown in the table below; the starting and scaling prices are both the total for 1× baseline TPM of input and output. Fine-tuned model prices are the same as the corresponding base model, subject to the actual console display.

MU Billing

Cost = Usage duration (hours) × Number of Model Units × Model Unit unit price

The "Model Unit unit price" takes the "Hourly unit price" column in the table below for postpaid scenarios; for monthly subscription prepaid billing, the formula becomes Number of monthly subscriptions × Number of Model Units × Monthly unit price.

• For the first month of a prepaid purchase, if you cancel the subscription early within the first month, the daily unit price (≈ Monthly unit price / 30) will be billed at 1.2 times the rate (less than one day is billed as one day)

NoteThe compute resources for the Model Unit postpaid method are first-come, first-served. If the purchase fails, a full refund will be issued.

Both MU and DTU use pre-paid billing. Early termination settles the used portion at a 1.2x coefficient for refund. For details, see Refund rules for configuration downgrades.

Supported Models and Pricing

DTU Price Table and Performance Baseline

China (Beijing) 2

Model	Spec	Baseline Input (TPM)	Baseline Output (TPM)	Starting Price (CNY/month)	Scaling Price (CNY/month)
qwen3.7-plus-2026-05-26	Fast	1,372,000	170,000	556,466	556,466
qwen3.6-27b	Standard	273,000	34,000	49,118	49,118
qwen3.5-397b-a17b	Standard	896,000	112,000	1,107,904	1,107,904
qwen3.5-122b-a10b	Fast	7,288,000	904,000	2,226,984	2,226,984
qwen3.5-35b-a3b	Standard	656,000	82,000	109,962	109,962
qwen3.5-27b	Standard	2,432,000	304,000	1,108,688	1,108,688
	Efficient	291,000	36,000	48,996	48,996
qwen3.5-4b	Standard	1,072,000	134,000	109,478	109,478
glm-5.2	Fast	1,004,000	120,000	1,895,260	1,895,260
glm-5.1	Standard	256,000	32,000	504,704	504,704
deepseek-v4-flash	Standard	2,240,000	280,000	1,107,400	1,107,400
deepseek-v4-flash-0731	Fast	1,676,000	213,000	277,984	277,984

The performance reference data for each model under standard workload is shown in the table below, subject to the actual console display.

The following performance reference data was measured at a 0% cache hit rate. In actual use, as the cache hit rate increases, model performance improves accordingly.

Model	Spec	Input Length	Output Length	Cache Hit Rate	First-token Latency (ms)	Per-token Latency (ms)
qwen3.7-plus-2026-05-26	Fast	16,000	2,000	0	2,418	15
qwen3.6-27b	Standard	4,000	500	0	1,292	19
qwen3.5-397b-a17b	Standard	4,000	500	0	996	27
qwen3.5-122b-a10b	Fast	16,000	2,000	0	568	8
qwen3.5-35b-a3b	Standard	4,000	500	0	471	10
qwen3.5-27b	Standard	4,000	500	0	703	14
	Efficient	4,000	500	0	1,448	23
qwen3.5-4b	Standard	4,000	500	0	552	6
glm-5.2	Fast	16,000	2,000	0	1,558	15
glm-5.1	Standard	4,000	500	0	769	27
deepseek-v4-flash	Standard	4,000	500	0	651	19
deepseek-v4-flash-0731	Fast	16,000	2,000	0	1,079	13

MU Price Table

Text generation

Qwen

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY) Minimum billing: minutes	Monthly Subscription Unit Price (CNY) Minimum billing: days
Qwen3.8-27B	qwen3.8-27b	MU9 x 4	¥204	¥98,400
Qwen3.7-Plus-2026-05-26	qwen3.7-plus-2026-05-26	MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
Qwen3.6-35B-A3B	qwen3.6-35b-a3b	MU1 x 8	¥432	¥208,944
		MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
		MU8 x 1	¥47	¥22,400
		MU9 x 1	¥51	¥24,600
Qwen3.6-27B	qwen3.6-27b	MU9 x 1	¥51	¥24,600
Qwen3.6-Flash-2026-04-16	qwen3.6-flash-2026-04-16	MU1 x 2	¥108	¥52,236
		MU3 x 8	¥1,096	¥527,752
Qwen3.6-Plus-2026-04-02	qwen3.6-plus-2026-04-02	MU1 x 8 MU1 x 16(PD disaggregation mode)	¥432 PD disaggregation mode:¥864	¥208,944 PD disaggregation mode:¥417,888
		MU2 x 8	¥504	¥240,288
Qwen3.5-397B-A17B	qwen3.5-397b-a17b	MU3 x 8 MU3 x 16(PD disaggregation mode)	¥1,096 PD disaggregation mode:¥2,192	¥527,752 PD disaggregation mode:¥1,055,504
		MU6 x 16	¥400	¥193,424
Qwen3.5-122B-A10B	qwen3.5-122b-a10b	MU1 x 4	¥216	¥104,472
		MU6 x 16	¥400	¥193,424
Qwen3.5-35B-A3B	qwen3.5-35b-a3b	MU1 x 2	¥108	¥52,236
		MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
		MU9 x 1	¥51	¥24,600
Qwen3.5-27B	qwen3.5-27b	MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
		MU8 x 1	¥47	¥22,400
		MU9 x 1	¥51	¥24,600
Qwen3.5-9B	qwen3.5-9b	MU1 x 2	¥108	¥52,236
		MU2 x 8	¥504	¥240,288
Qwen3.5-Flash-2026-02-23	qwen3.5-flash-2026-02-23	MU1 x 2	¥108	¥52,236
Qwen3.5-Plus-2026-02-15	qwen3.5-plus-2026-02-15	MU1 x 8 MU1 x 16(PD disaggregation mode)	¥432 PD disaggregation mode:¥864	¥208,944 PD disaggregation mode:¥417,888
		MU2 x 8	¥504	¥240,288
		MU3 x 8 MU3 x 16(PD disaggregation mode)	¥1,096 PD disaggregation mode:¥2,192	¥527,752 PD disaggregation mode:¥1,055,504
Qwen3-235B-A22B-Instruct-2507	qwen3-235b-a22b-instruct-2507	MU1 x 4	¥216	¥104,472
		MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
Qwen3-32B	qwen3-32b	MU6 x 16	¥400	¥193,424
Qwen3-30B-A3B-Thinking-2507	qwen3-30b-a3b-thinking-2507	MU1 x 2	¥108	¥52,236
Qwen3-4B	qwen3-4b	MU1 x 2	¥108	¥52,236
		MU5 x 1	¥21	¥10,139
Qwen3-Embedding-0.6B	qwen3-embedding-0.6b	MU5 x 1	¥21	¥10,139
		MU6 x 1	¥25	¥12,089
Qwen3-MoE-Rerank-0.6B	qwen3-moe-rerank-0.6b	MU5 x 1	¥21	¥10,139
Qwen3-Rerank-0.6B	qwen3-rerank-0.6b	MU5 x 1	¥21	¥10,139
		MU6 x 1	¥25	¥12,089
Qwen3-Max-2025-09-23	qwen3-max-2025-09-23	MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
Qwen3-Rerank	qwen3-rerank	MU5 x 1	¥21	¥10,139
Qwen2.5-Open-Source-72B	qwen2.5-72b-instruct	MU1 x 8	¥432	¥208,944
Qwen2.5-Open-Source-14B	qwen2.5-14b-instruct	MU1 x 2	¥108	¥52,236
Qwen2.5-Open-Source-7B	qwen2.5-7b-instruct	MU1 x 2	¥108	¥52,236
		MU5 x 1	¥21	¥10,139
Qwen-Plus-2025-07-28	qwen-plus-2025-07-28	MU1 x 4 MU1 x 16(PD disaggregation mode)	¥216 PD disaggregation mode:¥864	¥104,472 PD disaggregation mode:¥417,888
Qwen-Plus-2025-12-01	qwen-plus-2025-12-01	MU1 x 4	¥216	¥104,472
Qwen-Plus-Character-2025-11-06	qwen-plus-character-2025-11-06	MU1 x 4	¥216	¥104,472

GLM

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY) Minimum billing: minutes	Monthly Subscription Unit Price (CNY) Minimum billing: days
GLM-5.1	glm-5.1	MU2 x 8	¥504	¥240,288
		MU3 x 16(PD disaggregation mode)	PD disaggregation mode:¥2,192	PD disaggregation mode:¥1,055,504
		MU6 x 16	¥400	¥193,424
GLM-5	glm-5	MU3 x 16(PD disaggregation mode)	PD disaggregation mode:¥2,192	PD disaggregation mode:¥1,055,504
GLM-4.7	glm-4.7	MU6 x 32(PD disaggregation mode)	PD disaggregation mode:¥800	PD disaggregation mode:¥386,848

DeepSeek

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY) Minimum billing: minutes	Monthly Subscription Unit Price (CNY) Minimum billing: days
DeepSeek-v4-Flash	deepseek-v4-flash	MU1 x 8	¥432	¥208,944
		MU3 x 8	¥1,096	¥527,752
DeepSeek-v3.2	deepseek-v3.2	MU2 x 16(PD disaggregation mode)	PD disaggregation mode:¥1,008	PD disaggregation mode:¥480,576

More models

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY) Minimum billing: minutes	Monthly Subscription Unit Price (CNY) Minimum billing: days
Kimi-K2.5	kimi-k2.5	MU2 x 8	¥504	¥240,288

Model type:

• Instruct - The model performs inference in non-thinking mode after deployment.
• Thinking - The model performs inference in thinking mode after deployment.

Model deployment type:

• PD disaggregation mode - Reduces first-token latency and increases throughput.

  For models deployed in this mode, during inference the first-token computation (Prefill) and the subsequent token computation (Decode) are split across different compute nodes for execution.

Multimodal

Qwen-VL

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY) Minimum billing: minutes	Monthly Subscription Unit Price (CNY) Minimum billing: days
Qwen3-VL-235B-A22B-Thinking	qwen3-vl-235b-a22b-thinking	MU1 x 8	¥432	¥208,944
		MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
Qwen3-VL-32B-Instruct	qwen3-vl-32b-instruct	MU2 x 8	¥504	¥240,288
		MU3 x 8	¥1,096	¥527,752
Qwen3-VL-8B-Instruct	qwen3-vl-8b-instruct	MU1 x 2	¥108	¥52,236
		MU5 x 1	¥21	¥10,139
Qwen3-VL-4B-Instruct	qwen3-vl-4b-instruct	MU1 x 2	¥108	¥52,236
Qwen3-VL-2B-Instruct	qwen3-vl-2b-instruct	MU5 x 1	¥21	¥10,139
Qwen3-VL-Embedding-2B	qwen3-vl-embedding-2b	MU5 x 1	¥21	¥10,139
Qwen3-VL-Flash-2025-10-15	qwen3-vl-flash-2025-10-15	MU1 x 4	¥216	¥104,472
Qwen3-VL-Plus-2025-09-23	qwen3-vl-plus-2025-09-23	MU1 x 4	¥216	¥104,472
Qwen-VL-Max-2025-08-13	qwen-vl-max-2025-08-13	MU6 x 4	¥100	¥48,356

Qwen Omni

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY) Minimum billing: minutes	Monthly Subscription Unit Price (CNY) Minimum billing: days
Qwen3.5-Omni-Flash	qwen3.5-omni-flash	MU8 x 1	¥47	¥22,400
		MU9 x 1	¥51	¥24,600

Model Type:

• Instruct - The model performs inference in non-thinking mode after deployment.
• Thinking - The model performs inference in thinking mode after deployment.
• Instruct/Thinking - You can choose whether to enable thinking mode when deploying the model.

Speech Synthesis

CosyVoice

Model Name	Model Code	Model Unit Specification	Hourly Unit Price (CNY)	Monthly Subscription Unit Price (CNY)
cosyvoice-v3-flash	cosyvoice-v3-flash	MU5	¥21	¥10,139

Deployment Creation

DTU Deployment Creation

Before using DTU, enable the DTU feature in the console and apply for a resource quota. Once enabled, select the target model on the Dedicated Deployment page of the Bailian console and choose DTU as the billing method to create a deployment.

WarningDTU deployment does not currently support creation and management via API. Please complete enablement, creation, scaling, and renewal in the Bailian console.

For the basic workflow of general model deployment, see Dedicated Deployment Overview.

The form fields for creating a deployment are shown in the table below.

Parameter	Description	Required	Value Description
Service Name	Name of the deployment service	Yes	Custom
Model	Target model to deploy	Yes	Dropdown selection
Deployment Template	Deployment architecture	No	Dropdown selection, default to the first
Payment Type	Billing method	Yes	Pre-paid (monthly)
Input Throughput Quota	Purchased input TPM capacity	Yes	Integer multiple of baseline input TPM (kTPM)
Output Throughput Quota	Purchased output TPM capacity	Yes	Integer multiple of baseline output TPM (kTPM)
Purchase Duration	Purchase period	Yes	1-12 (integer, months)
Auto Renewal	Auto-renew on expiry	No	On/Off
Single Renewal Duration	Required when auto-renewal is enabled	Yes	1-12 (integer, months)

MU Deployment Configuration

Model Unit

Configuration item	Configuration details
Service name	A custom name for the deployment service.
Model	Select the model to deploy, including platform preset models and fine-tuned models.
Model unit type	Select the deployment specification. Different specifications correspond to different computing power and performance.
Replica count	Set the initial number of deployment replicas, which affects the concurrent processing capability of the service.
Deployment template	Select a deployment template (for example, "single-node deployment"). Different templates correspond to different resource configuration schemes. Available only in the model unit billing mode.
Model inference mode	For some models, when deployed inModel Unit mode, you can configure the inference mode, maximum context, and more. Instruct - The model performs inference in non-thinking mode after deployment. Thinking - The model performs inference in thinking mode after deployment.
Maximum context	TheModel Unit deployment mode of some models supports this setting. The maximum context length depends on the model type.
Service throttling	TheModel Unit deployment mode of some models supports this setting, which can limit the RPM and TPM of model calls.

Capacity Planning

DTU is sold by input/output TPM; the purchased TPM defines the maximum tokens processable per minute. Capacity planning aims to derive the purchase multiple from your business's peak token demand, then verify through load testing that the actually sustainable concurrency and latency meet requirements. Baseline input/output TPM and standard-workload performance references for each model are in the tables above.

Use a load-testing tool such as evalscope to test against your real business scenario (input/output length, concurrency, latency requirements), then determine the purchase multiple against the pricing table.

Capacity Planning Method

1. Profile peak business metrics: typical request input/output token length, target concurrency, and requirements for first-token latency (TTFT) and per-token generation latency (TPOT).
2. Estimate peak throughput demand: peak input TPM ≈ concurrency × per-request input length ÷ per-request processing time (minutes); peak output TPM ≈ concurrency × per-request output length ÷ per-request generation time (minutes).
3. Compute the purchase multiple: input multiple = ⌈peak input TPM ÷ baseline input TPM⌉, output multiple = ⌈peak output TPM ÷ baseline output TPM⌉. DTU requires input and output to scale together, so take the larger of the two as the final multiple.
4. Verify with load testing: after purchasing the computed multiple, re-test with evalscope under the target workload to confirm actual throughput and latency meet business requirements. If latency is high, raise concurrency within the TPM headroom to improve effective throughput.

If latency requirements are not strict, you can raise concurrency — within the purchased TPM cap — to increase actual throughput.

Business estimation example:

For a model with baseline input 656K TPM and baseline output 82K TPM: business peak profiling shows peak input ~1,300K TPM and peak output ~160K TPM. Compute multiples: input ⌈1300÷656⌉=2, output ⌈160÷82⌉=2; take the larger (2×), purchasing input 1,312K TPM (656K×2) + output 164K TPM (82K×2). Then load-test with evalscope at the peak workload to confirm latency passes. If business volume doubles (peak input ~2,600K, output ~320K), multiples become input ⌈2600÷656⌉=4, output ⌈320÷82⌉=4; purchase 4×: input 2,624K TPM (656K×4) + output 328K TPM (82K×4), and re-test.

Estimation formula:

Purchase multiple = max(⌈peak input TPM ÷ baseline input TPM⌉, ⌈peak output TPM ÷ baseline output TPM⌉)

Scaling, Renewal & Unsubscription

DTU Scaling, Renewal & Unsubscription

The deployment list shows the purchased input/output TPM quotas; the details page shows the TPM capacity details.

Scaling: Click "Scale" in the deployment list to modify the input/output TPM capacity. Scaling up shows the additional amount due; scaling down shows the estimated refund. Input and output must be increased or decreased together; one cannot increase while the other decreases. Only running deployments can be operated.

Renewal: Pre-paid monthly deployments can be renewed upon expiry. The renewal button on the details page enters the renewal flow, supporting an auto-renewal toggle and a single renewal duration.

Unsubscription: Prepaid monthly orders support early termination; the used portion is settled at a 1.2x coefficient for refund. For details, see Refund rules for configuration downgrades. Unsubscription is irreversible; after unsubscription, dedicated resources are released and the service stops.

MU Scaling and Renewal

• Model unit (billed by duration): Click the Scaling button to self-service, manually adjust the number of instances (replica count). You can also configure an auto-scaling policy (including scaling thresholds, minimum/maximum replica count, scheduled scaling, etc.) via the scaling configuration button in the operation column.
• Renewal: Prepaid monthly services can be renewed to extend the service time; auto-renewal is supported.
• Unsubscription: If you cancel early within the first month of a prepaid purchase, the daily unit price is billed at 1.2 times the rate. For details, see Refund rules for configuration downgrades.

FAQ

What is the difference between DTU deployment and Token pay-as-you-go billing?

Token pay-as-you-go bills by token usage on a shared resource pool; DTU uses dedicated GPU resources and bills by input/output TPM, suitable for scenarios requiring stable throughput and dedicated deployment.

What billing methods does DTU deployment support?

Only pre-paid (monthly) billing is supported.

How do I view the usage of purchased TPM?

You can view the purchased input/output TPM quotas in the deployment list and details page of Dedicated Deployment in the Bailian console. See Dedicated Deployment Overview.