ModelScope API-Inference

LLM 5.7/10 Jun 16, 2026

ModelScope, Alibaba's open model community, provides free API-Inference to registered users across 900+ trending models including Qwen LLMs and image-generation models.

Requests / day: 2K
Card required: No

ModelScope API-Inference scores 5.7/10 for free value in LLM. The main constraint: Reported 2,000 requests/day total per user (with dynamic per-model caps up to ~500/day) and dynamic concurrency, free for registered users. Requires Alibaba Cloud account binding plus real-name verification.

Category LLM Kind Hosted LLM/multimodal inference API (free quota) Caveats Requires linking an Alibaba Cloud account and completing real-name verification (eligibility/identity gate); quotas are dynamic and can change; primarily China-region oriented. Renewal daily Card required No Eligibility anyone, region restricted, approval required

Free value score

5.7 / 10

Show the math

Model Quality 9.2/10

value = 51 artificial_analysis_score

weighted score = 0.4 * 9.2/10

Request Rate 1.3/10

value = 1.39 req/min

weighted score = 0.2 * 1.3/10

Requests 4.6/10

value = 60000 req/mo

weighted score = 0.2 * 4.6/10

Access 4/10

value = 4/10

weighted score = 0.2 * 4/10

Key metrics

ZhipuAI/GLM-5.2 51 Best model

2.0K R/day Requests/day

Yes Commercial use no published noncom clause

Models

40 free · 57+ total

Limits may be stale — last re-verified outside this section's freshness window.

Model	Family	AA	Access
ZhipuAI/GLM-5.2	GLM	51.1	free
MiniMax/MiniMax-M3	MiniMax	44.4	free
deepseek-ai/DeepSeek-V4-Pro	DeepSeek	44.3	free
deepseek-ai/DeepSeek-V4-Flash	DeepSeek	40.3	free
ZhipuAI/GLM-5.1	GLM	40.2	free
ZhipuAI/GLM-5	GLM	39.5	free
MiniMax/MiniMax-M2.7	MiniMax	38.1	free
moonshotai/Kimi-K2.5	Kimi	38.1	free
Qwen/Qwen3.5-27B	Qwen	33.8	free
ZhipuAI/GLM-4.7-Flash	GLM	33.8	free
Qwen/Qwen3.5-122B-A10B	Qwen	32.3	free
Qwen/Qwen3.5-397B-A17B	Qwen	32	free
Qwen/Qwen3.5-35B-A3B	Qwen	29.3	free
deepseek-ai/DeepSeek-V3.2	DeepSeek	24.7	free
LLM-Research/Llama-4-Maverick-17B-128E-Instruct	Llama	14.3	free
LLM-Research/c4ai-command-r-plus-08-2024	Command-R	—	free
MiniMax/MiniMax-M1-80k	MiniMax	—	free
MiniMax/MiniMax-M2.5	MiniMax	—	free
mistralai/Ministral-8B-Instruct-2410	Mistral	—	free
mistralai/Mistral-Large-Instruct-2407	Mistral	—	free
mistralai/Mistral-Small-Instruct-2409	Mistral	—	free
MusePublic/Qwen-Image-Edit	Qwen	—	free
PaddlePaddle/ERNIE-4.5-300B-A47B-PT	ERNIE	—	free
PaddlePaddle/ERNIE-4.5-VL-28B-A3B-PT	ERNIE	—	free
Qwen/Qwen-Image-Edit	Qwen	—	free
Qwen/Qwen3-14B	Qwen	—	free
Qwen/Qwen3-235B-A22B	Qwen	—	free
Qwen/Qwen3-235B-A22B-Instruct-2507	Qwen	—	free
Qwen/Qwen3-235B-A22B-Thinking-2507	Qwen	—	free
Qwen/Qwen3-30B-A3B	Qwen	—	free
Qwen/Qwen3-30B-A3B-Thinking-2507	Qwen	—	free
Qwen/Qwen3-32B	Qwen	—	free
Qwen/Qwen3-4B	Qwen	—	free
Qwen/Qwen3-8B	Qwen	—	free
Qwen/Qwen3-Coder-30B-A3B-Instruct	Qwen	—	free
Qwen/Qwen3-Next-80B-A3B-Instruct	Qwen	—	free
Qwen/Qwen3-Next-80B-A3B-Thinking	Qwen	—	free
Qwen/Qwen3-VL-235B-A22B-Instruct	Qwen	—	free
Qwen/Qwen3-VL-8B-Instruct	Qwen	—	free
Qwen/Qwen3-VL-8B-Thinking	Qwen	—	free

Showing first 40 of a larger catalog.

Compare all free models across providers →

What you get

Reported 2,000 free API calls/day across 900+ models

Includes Qwen LLMs, multimodal models, and AIGC image generation

OpenAI-SDK-compatible endpoint (api-inference.modelscope.cn/v1)

Free quota separate from any paid Alibaba Cloud usage

Gotchas

Alibaba Cloud account + real-name verification

Users must link an Alibaba Cloud account and complete government-ID real-name (KYC) verification before the free quota is activated.
China-region oriented service

The platform and its documentation are primarily in Chinese and served from China-region infrastructure, creating friction for international developers.

Sources

"Free quota: Each user 2,000 API calls per day (*subject to the latest rules on the official website) · Quota reset: Automatically resets daily at 00:00 UTC+8, does not support cross-day accumulation or rollover"
https://docs.cherry-ai.com/docs/en-us/pre-basic/providers/modelscope
"API-Inference is free for registered users. Current published limits are 2,000 requests/day per user (total across models), with per-model daily quotas dynamically adjusted and capped at 500; concurrency is also dynamically rate-limited. Requires Alibaba Cloud account binding and real-name verification"
https://github.com/mnfst/awesome-free-llm-apis/blob/main/README.md
"Kimi K2 Thinking (1T): Released November 4, 2025; Intelligence Index: 67 ... Kimi K2 Thinking is arguably the best open model in the world by benchmark score. Artificial Analysis currently ranks it as the strongest model not made by OpenAI, Google, or Anthropic."
https://www.understandingai.org/p/the-best-chinese-open-weight-models
"When attempting to inference any model using the API, I receive an error message stating that I need to bind my Alibaba Cloud account."
https://github.com/modelscope/modelscope/issues/1524

llm-api china qwen image-generation free-quota

Open service Source ← All LLM

Alternatives

Same category · similar quota

OpenRouterOpenRouter · freemium · LLM router

5.7

Best first BYOK option: one API for many free and paid models.

37.8 AA20/min

May be outdatedChecked 29d ago

Fireworks AIFireworks AI · finite_credit · Fast hosted open-model LLM API

5.6

Fireworks AI provides fast, production-grade inference on open models (Llama, DeepSeek, Qwen, gpt-oss) with $1 in free credits for new accounts.

$1.00 one-time*51.1 AA519.5K tok

May be outdatedChecked 28d ago

One-time

BytePlus ModelArk Coding LiteBytePlus ModelArk Coding Lite · paid · Paid plan

5.5

~1,900 requests/5h (~24,000/mo) across Seed-2.0-Pro/Lite/Code, Kimi K2.5, GLM-5.1, DeepSeek V3.2; OpenAI endpoint for all models + Anthropic endpoint (subset) in Claude Code/OpenCode; Singapore region

1×40.2 AA1.0×

May be outdatedChecked 23d ago

Paid onlyMay train