[Feature]: Add Huawei Cloud ModelArts MaaS as a new LLM provider"

[netoxp70]

Check for existing issues

• I have searched the existing issues and checked that my issue is not a duplicate.

The Feature

The Feature

Add support for Huawei Cloud ModelArts MaaS (Model as a Service) as a new LLM provider in LiteLLM.

Huawei Cloud MaaS exposes an OpenAI-compatible chat completions endpoint, so integration is pretty straightforward:

• Default endpoint: https://api-ap-southeast-1.modelarts-maas.com/openai/v1/chat/completions
• Auth: Bearer token via Authorization header
• Docs: Huawei MaaS docs

Models available

Model	Context	Reasoning	Tool Calling
DeepSeek-V3	32K	❌	✅
DeepSeek-V4-Pro	1M	✅	❌
DeepSeek-V4-Flash	1M	✅	✅
deepseek-v3.1-terminus	32K	❌	✅
deepseek-v3.2	32K	❌	✅
deepseek-r1-250528	32K	✅	❌
DeepSeek-V3-32K/64K/128K	32K–128K	❌	✅
DeepSeek-R1-32K/64K/128K	32K–128K	✅	❌
DeepSeek-R1-distil-Qwen-32B	32K	✅	❌
DeepSeek-R1-distill-LLama-70B/8B	32K	✅	❌
Qwen3-235B-A22B	32K	✅	✅
Qwen3-32B, Qwen3-30B-A3B, Qwen3-14B, Qwen3-8B	32K	✅	✅
Qwen2.5-72B	32K	❌	✅
Qwen2.5-VL-32B	32K	❌	👁️ Vision
QWQ-32B	32K	✅	❌
glm-5, glm-5.1	32K	❌	✅

Motivation

Huawei Cloud is a pretty big deal in Asia-Pacific and enterprise settings. ModelArts MaaS gives you access to solid models like DeepSeek, Qwen3, and GLM at decent prices. A bunch of LiteLLM users working in regions where Huawei Cloud is popular (AP-Southeast, China, etc.) can't use these models through LiteLLM right now.

Since the API is OpenAI-compatible, the integration follows the same pattern as existing providers like OVHCloud and DashScope, so the implementation stays minimal and easy to maintain.

What part of LiteLLM is this about?

SDK (the litellm Python package)

Implementation scope

• litellm/llms/huawei_cloud/chat/transformation.py — HuaweiCloudChatConfig extending OpenAIGPTConfig
• litellm/llms/huawei_cloud/utils.py — HuaweiCloudException
• Registration in types/utils.py (LlmProviders enum), constants.py, __init__.py, main.py, utils.py, _lazy_imports_registry.py, get_llm_provider_logic.py
• 28 models added to model_prices_and_context_window.json with accurate pricing and capability flags
• UI: provider_info_helpers.tsx updated

Usage (after integration)

import litellm

response = litellm.completion(
    model="huawei_cloud/DeepSeek-V3",
    messages=[{"role": "user", "content": "Hello!"}],
    api_key="<HUAWEI_CLOUD_API_KEY>",
)

### Motivation, pitch

Huawei Cloud ModelArts MaaS is a widely used LLM provider in the Asia-Pacific region and enterprise environments. The platform exposes state-of-the-art models (DeepSeek, Qwen3, GLM) via an OpenAI-compatible API with competitive pricing.

Currently, LiteLLM users operating in regions where Huawei Cloud is available (AP-Southeast, China, etc.) cannot use these models through the library. The integration is straightforward since it follows the same pattern as providers like OVHCloud and Dashscope (OpenAI-compatible), keeping the code minimal and maintainable.

Use case
Organizations already using Huawei Cloud as their infrastructure provider who need a unified interface for multiple LLM providers via LiteLLM.

### What part of LiteLLM is this about?

SDK (litellm Python package)

### LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?

No

### Twitter / LinkedIn details

https://www.linkedin.com/in/milton-neto-46aa51a9/

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.