ModelScope

ModelScope ist Alibabas öffentlicher Modell-Hub. Die Inferenz-API hält einen großen Gratis-Katalog bereit — 50 Modelle auf der Live-Liste, darunter MiniMax M3, Qwen 3.8, DeepSeek V4 und GLM — bei 2.000 Anfragen pro Tag, OpenAI-kompatibel.

API-Schlüssel holen

Die Anmeldung will eine Telefonnummer — oft eine chinesische Festland-Nummer —, bevor der Inferenz-Key funktioniert. 1M Kontext bei den langen Modellen. Wer Qwen direkt von der Quelle will, dem steht Alibaba Model Studio offen; ModelScope ist der breitere Zoo.

Kostenlose Modelle: 50
Max. Kontext: 1M
Gratis-Tarif: 2.000 Anfragen/Tag
Anforderung: Telefonverifizierung, keine Karte

Endpunkt

Basis-URLhttps://api-inference.modelscope.cn/v1

UmgebungsvariableMODELSCOPE_API_KEY

Ratengrenze2,000 requests/day across all models

OpenAI-kompatibel
Text
Bild
Audio
Video
Vision
Reasoning

Codebeispiele

npm install openai

import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api-inference.modelscope.cn/v1',
  apiKey: process.env.MODELSCOPE_API_KEY
})

const response = await client.chat.completions.create({
  model: 'MiniMax/MiniMax-M3',
  messages: [{ role: 'user', content: 'Explain closures in one paragraph.' }]
})

console.log(response.choices[0].message.content)

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api-inference.modelscope.cn/v1",
    api_key=os.environ["MODELSCOPE_API_KEY"],
)

response = client.chat.completions.create(
    model="MiniMax/MiniMax-M3",
    messages=[{"role": "user", "content": "Explain closures in one paragraph."}],
)

print(response.choices[0].message.content)

curl https://api-inference.modelscope.cn/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MODELSCOPE_API_KEY" \
  -d '{
    "model": "MiniMax/MiniMax-M3",
    "messages": [{"role": "user", "content": "Explain closures in one paragraph."}]
  }'

Wissenswert

Für manche Kontotypen verlangt die Registrierung eine chinesische Festland-Telefonnummer.

Modelle, die Sie gratis bekommen50

DeepSeek V4 Flashdeepseek-ai/DeepSeek-V4-Flash-0731
DeepSeek V4 Prodeepseek-ai/DeepSeek-V4-Pro
DeepSeek V4 Pro 0813deepseek-ai/DeepSeek-V4-Pro-0813
EA 29B A4Bearly-access/EA-29B-A4B
AntAngelMedMedAIBase/AntAngelMed
LongCat Flash Litemeituan-longcat/LongCat-Flash-Lite
MiniMax M1MiniMax/MiniMax-M1-80k
MiniMax M3MiniMax/MiniMax-M3
Mistral Largemistralai/Mistral-Large-Instruct-2407
Qwen Image EditMusePublic/Qwen-Image-Edit
CompassJudger 32Bopencompass/CompassJudger-1-32B-Instruct
InternVL 3.5OpenGVLab/InternVL3_5-241B-A28B
ERNIE 4.5 0.3BPaddlePaddle/ERNIE-4.5-0.3B-PT
ERNIE 4.5 21BPaddlePaddle/ERNIE-4.5-21B-A3B-PT
ERNIE 4.5 300BPaddlePaddle/ERNIE-4.5-300B-A47B-PT
ERNIE 4.5 VL 28BPaddlePaddle/ERNIE-4.5-VL-28B-A3B-PT
Qwen Image EditQwen/Qwen-Image-Edit
Qwen 3 14BQwen/Qwen3-14B
Qwen 3 235BQwen/Qwen3-235B-A22B
Qwen 3 235BQwen/Qwen3-235B-A22B-Instruct-2507
Qwen 3 235B ThinkingQwen/Qwen3-235B-A22B-Thinking-2507
Qwen 3 30B A3BQwen/Qwen3-30B-A3B
Qwen 3 30B A3B ThinkingQwen/Qwen3-30B-A3B-Thinking-2507
Qwen 3 4BQwen/Qwen3-4B
Qwen 3 8BQwen/Qwen3-8B
Qwen 3 Coder 30BQwen/Qwen3-Coder-30B-A3B-Instruct
Qwen 3 Embedding 0.6BQwen/Qwen3-Embedding-0.6B
Qwen 3 Embedding 4BQwen/Qwen3-Embedding-4B
Qwen 3 Embedding 8BQwen/Qwen3-Embedding-8B
Qwen 3 Next 80BQwen/Qwen3-Next-80B-A3B-Instruct
Qwen 3 Next 80B ThinkingQwen/Qwen3-Next-80B-A3B-Thinking
Qwen 3 VL 235BQwen/Qwen3-VL-235B-A22B-Instruct
Qwen 3 VL 8BQwen/Qwen3-VL-8B-Instruct
Qwen 3 VL 8B ThinkingQwen/Qwen3-VL-8B-Thinking
Qwen 3.5 122BQwen/Qwen3.5-122B-A10B
Qwen 3.5 27BQwen/Qwen3.5-27B
Qwen 3.5 35BQwen/Qwen3.5-35B-A3B
Qwen 3.5 397BQwen/Qwen3.5-397B-A17B
Qwen 3.8 27BQwen/Qwen3.8-27B
Qwen 3.8 Flash NextQwen/Qwen3.8-Flash-Next
Intern S1Shanghai_AI_Laboratory/Intern-S1
Intern S1 MiniShanghai_AI_Laboratory/Intern-S1-mini
Intern S2Shanghai_AI_Laboratory/Intern-S2-Preview
Step 3.5 Flashstepfun-ai/Step-3.5-Flash
Step 3.7 Flashstepfun-ai/Step-3.7-Flash
Hunyuan Hy3Tencent-Hunyuan/Hy3
XiYanSQL 32BXGenerationLab/XiYanSQL-QwenCoder-32B-2412
XiYanSQL 32B 2504XGenerationLab/XiYanSQL-QwenCoder-32B-2504
GLM 4.7 FlashZhipuAI/GLM-4.7-Flash
GLM 5.2ZhipuAI/GLM-5.2