跳到主要内容
非官方简体中文翻译，内容版权归 Mastra 所有，遵循 Apache-2.0 许可。

Hugging Face logoHugging Face

Access 82 Hugging Face models through Mastra's model router. Authentication is handled automatically using the HF_TOKEN environment variable.

Learn more in the Hugging Face documentation.

.env
HF_TOKEN=your-api-token
src/mastra/agents/my-agent.ts
import { Agent } from "@mastra/core/agent";

const agent = new Agent({
  id: "my-agent",
  name: "My Agent",
  instructions: "You are a helpful assistant",
  model: "huggingface/MiniMaxAI/MiniMax-M1-80k"
});

// Generate a response
const response = await agent.generate("Hello!");

// Stream a response
const stream = await agent.stream("Tell me a story");
for await (const chunk of stream) {
  console.log(chunk);
}
备注

Mastra uses the OpenAI-compatible /chat/completions endpoint. Some provider-specific features may not be available. Check the Hugging Face documentation for details.

Models
Models的直接链接

ModelContextToolsReasoningImageAudioVideoInput $/1MOutput $/1M
huggingface/deepseek-ai/DeepSeek-R164K$0.70$3
huggingface/deepseek-ai/DeepSeek-R1-0528164K$3$5
huggingface/deepseek-ai/DeepSeek-V364K$0.40$1
huggingface/deepseek-ai/DeepSeek-V3-0324164K$0.27$1
huggingface/deepseek-ai/DeepSeek-V3.1131K$0.27$1
huggingface/deepseek-ai/DeepSeek-V3.2164K$0.28$0.40
huggingface/deepseek-ai/DeepSeek-V4-Flash1.0M$0.14$0.28
huggingface/deepseek-ai/DeepSeek-V4-Flash-07311.0M$0.14$0.28
huggingface/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp1.0M$0.44$1
huggingface/deepseek-ai/DeepSeek-V4-Pro1.0M$0.43$0.87
huggingface/deepseek-ai/DeepSeek-V4-Pro-08131.0M$1$4
huggingface/deepseek-ai/DeepSeek-V4.1-Flash1.0M$0.30$1
huggingface/deepseek-ai/DeepSeek-V4.1-Flash-10171.0M$0.30$1
huggingface/google/gemma-3-12b-it131K$0.05$0.15
huggingface/google/gemma-3-27b-it131K$0.08$0.16
huggingface/google/gemma-3-4b-it131K$0.05$0.10
huggingface/google/gemma-4-26B-A4B-it262K$0.13$0.40
huggingface/google/gemma-4-31B-it262K$0.14$0.40
huggingface/meta-llama/Llama-3.1-8B-Instruct131K$0.06$0.06
huggingface/meta-llama/Llama-3.3-70B-Instruct131K$0.59$0.79
huggingface/MiniMaxAI/MiniMax-M1-80k1.0M$0.55$2
huggingface/MiniMaxAI/MiniMax-M2205K$0.30$1
huggingface/MiniMaxAI/MiniMax-M2.1205K$0.30$1
huggingface/MiniMaxAI/MiniMax-M2.5205K$0.30$1
huggingface/MiniMaxAI/MiniMax-M2.7205K$0.30$1
huggingface/MiniMaxAI/MiniMax-M3524K$0.30$1
huggingface/moonshotai/Kimi-K2-Instruct131K$1$3
huggingface/moonshotai/Kimi-K2-Instruct-0905262K$1$3
huggingface/moonshotai/Kimi-K2-Thinking262K$0.60$3
huggingface/moonshotai/Kimi-K2.5262K$0.60$3
huggingface/moonshotai/Kimi-K2.6262K$0.95$4
huggingface/moonshotai/Kimi-K2.7-Code262K$0.95$4
huggingface/moonshotai/Kimi-K31.0M$3$15
huggingface/openai/gpt-oss-120b131K$0.25$0.69
huggingface/openai/gpt-oss-20b131K$0.10$0.50
huggingface/Qwen/Qwen2.5-Coder-32B-Instruct131K$0.06$0.20
huggingface/Qwen/Qwen3-235B-A22B41K$0.20$0.80
huggingface/Qwen/Qwen3-235B-A22B-Instruct-2507262K$0.85$3
huggingface/Qwen/Qwen3-235B-A22B-Thinking-2507262K$0.30$3
huggingface/Qwen/Qwen3-30B-A3B41K$0.12$0.50
huggingface/Qwen/Qwen3-32B131K$0.29$0.59
huggingface/Qwen/Qwen3-Coder-30B-A3B-Instruct262K$0.07$0.26
huggingface/Qwen/Qwen3-Coder-480B-A35B-Instruct262K$2$2
huggingface/Qwen/Qwen3-Coder-Next262K$0.20$2
huggingface/Qwen/Qwen3-Embedding-4B32K$0.01—
huggingface/Qwen/Qwen3-Embedding-8B32K$0.01—
huggingface/Qwen/Qwen3-Next-80B-A3B-Instruct262K$0.25$1
huggingface/Qwen/Qwen3-Next-80B-A3B-Thinking262K$0.30$2
huggingface/Qwen/Qwen3-VL-235B-A22B-Instruct131K$0.30$2
huggingface/Qwen/Qwen3-VL-235B-A22B-Thinking131K$0.98$4
huggingface/Qwen/Qwen3-VL-30B-A3B-Instruct131K$0.20$0.70
huggingface/Qwen/Qwen3.5-122B-A10B262K$0.40$3
huggingface/Qwen/Qwen3.5-27B262K$0.30$2
huggingface/Qwen/Qwen3.5-35B-A3B262K$0.25$2
huggingface/Qwen/Qwen3.5-397B-A17B262K$0.60$4
huggingface/Qwen/Qwen3.5-9B262K$0.17$0.25
huggingface/Qwen/Qwen3.6-27B262K$0.47$3
huggingface/Qwen/Qwen3.6-35B-A3B262K$0.15$0.95
huggingface/Qwen/Qwen3.8-2.4T-A95B262K$3$6
huggingface/Qwen/Qwen3.8-27B262K$0.40$3
huggingface/stepfun-ai/Step-3.5-Flash262K$0.10$0.30
huggingface/stepfun-ai/Step-3.7-Flash262K$0.20$1
huggingface/tencent/Hy3262K$0.14$0.58
huggingface/tencent/Hy4-preview1.0M$0.83$3
huggingface/thinkingmachines/Inkling1.0M$1$4
huggingface/thinkingmachines/Inkling-Small524K$0.50$1
huggingface/XiaomiMiMo/MiMo-V2-Flash262K$0.10$0.30
huggingface/XiaomiMiMo/MiMo-V2.5262K$0.40$2
huggingface/XiaomiMiMo/MiMo-V2.5-Pro1.0M$1$3
huggingface/zai-org/AutoGLM-Phone-9B-Multilingual66K$0.04$0.14
huggingface/zai-org/GLM-4.5131K$0.60$2
huggingface/zai-org/GLM-4.5-Air131K$0.13$0.85
huggingface/zai-org/GLM-4.5V66K$0.60$2
huggingface/zai-org/GLM-4.6205K$0.55$2
huggingface/zai-org/GLM-4.6V-Flash131K$0.30$0.90
huggingface/zai-org/GLM-4.7205K$0.60$2
huggingface/zai-org/GLM-4.7-Flash200K——
huggingface/zai-org/GLM-5203K$1$3
huggingface/zai-org/GLM-5.1203K$1$3
huggingface/zai-org/GLM-5.2262K$1$4
huggingface/zai-org/GLM-5.31.0M$1$4
huggingface/zai-org/GLM-5.3-Flash1.0M$0.15$0.50
82 available models

Model availability, capabilities, context windows, and pricing are sourced from models.dev and may change.

Advanced configuration
Advanced configuration的直接链接

Custom headers
Custom headers的直接链接

src/mastra/agents/my-agent.ts
const agent = new Agent({
  id: "custom-agent",
  name: "custom-agent",
  model: {
    url: "https://router.huggingface.co/v1",
    id: "huggingface/MiniMaxAI/MiniMax-M1-80k",
    apiKey: process.env.HF_TOKEN,
    headers: {
      "X-Custom-Header": "value"
    }
  }
});

Dynamic model selection
Dynamic model selection的直接链接

src/mastra/agents/my-agent.ts
const agent = new Agent({
  id: "dynamic-agent",
  name: "Dynamic Agent",
  model: ({ requestContext }) => {
    const useAdvanced = requestContext.task === "complex";
    return useAdvanced
      ? "huggingface/zai-org/GLM-5.3-Flash"
      : "huggingface/MiniMaxAI/MiniMax-M1-80k";
  }
});
On this page