vercel/ai · error · NoSuchModelError

No such embeddingModel: ${modelId}

Error message

No such embeddingModel: ${modelId}

What it means

The z.ai (Zhipu) provider does not implement embedding models; its embeddingModel/textEmbeddingModel factories unconditionally throw NoSuchModelError with modelType 'embeddingModel'. Any attempt to obtain an embedding model from @ai-sdk/zai triggers this, since the provider only supports language models.

Source

Thrown at packages/zai/src/zai-provider.ts:93

      `ai-sdk/zai/${VERSION}`,
    );

  const createLanguageModel = (modelId: ZaiChatModelId) =>
    new ZaiChatLanguageModel(modelId, {
      provider: 'zai.chat',
      baseURL,
      headers: getHeaders,
      fetch: options.fetch,
    });

  const provider = (modelId: ZaiChatModelId) => createLanguageModel(modelId);

  provider.specificationVersion = 'v4' as const;
  provider.languageModel = createLanguageModel;
  provider.chat = createLanguageModel;

  provider.embeddingModel = (modelId: string) => {
    throw new NoSuchModelError({ modelId, modelType: 'embeddingModel' });
  };
  provider.textEmbeddingModel = provider.embeddingModel;
  provider.imageModel = (modelId: string) => {
    throw new NoSuchModelError({ modelId, modelType: 'imageModel' });
  };

  return provider;
}

export const zai = createZai();

View on GitHub (pinned to 69428b1f8b)

Solutions

1. Use a provider with embedding support (e.g. @ai-sdk/openai textEmbeddingModel) for embed/embedMany.
2. Keep zai for chat/language models only and split your model configuration by capability.
3. Add an explicit capability check or registry so embedding requests never route to zai.

Example fix

// before
const { embedding } = await embed({ model: zai.textEmbeddingModel('embedding-3'), value });

// after
import { createOpenAI } from '@ai-sdk/openai';
const openai = createOpenAI();
const { embedding } = await embed({ model: openai.textEmbeddingModel('text-embedding-3-small'), value });

const ZAI_SUPPORTS_EMBEDDINGS = false;
function getEmbeddingModel(preferred: 'zai' | 'openai') {
  return preferred === 'zai' && !ZAI_SUPPORTS_EMBEDDINGS
    ? openai.textEmbeddingModel('text-embedding-3-small')
    : /* other */ null;
}

function supportsEmbeddings(provider: any): boolean {
  return typeof provider?.textEmbeddingModel === 'function' && provider.specificationVersion != null &&
    !/NoSuchModelError/.test(provider.textEmbeddingModel.toString());
}

try {
  return await embed({ model: zai.textEmbeddingModel(id), value });
} catch (e) {
  if (NoSuchModelError.isInstance(e)) {
    return await embed({ model: openai.textEmbeddingModel('text-embedding-3-small'), value });
  }
  throw e;
}

When it happens

Trigger: zai.embeddingModel('embedding-3') or zai.textEmbeddingModel('...') called directly; embed()/embedMany() with model: zai('some-model'); generic provider-selection code mapping 'zai' to an embedding request.

Common situations: Assuming every AI SDK provider supports embeddings; migrating an embedding pipeline from openai/mistral to zai; capability-agnostic model factory code in multi-provider apps.

Related errors

• AI_NoSuchModelError
• embeddingModel
• ElevenLabs does not provide embedding models
• AI_NoSuchModelError
• No such embeddingModel: ${modelId}

AI-assisted analysis of vercel/ai@69428b1f8b (2026-08-30). Data as JSON: /api/errors/1790d142ab564b96. Report an issue: GitHub.