Sherlock AI models on CloudFerro Cloud

Sherlock AI supports inference with large language models, models with vision capabilities, and embedding models. When calling the API, always pass the exact value from the Model ID column in the model field of your request.

Model availability can change. Use the models endpoint to confirm which model IDs are currently available in your project before building an integration.

Use a model ID in API requests

The selected model is specified in the request body. For example:

{
  "model": "speakleash/Bielik-11B-v3.0-Instruct",
  "messages": [
    {
      "role": "user",
      "content": "Explain what Sherlock AI is."
    }
  ]
}

Use the model ID exactly as shown in the tables below. Model IDs are case-sensitive.

Large language models

Table 7 Large language models
Provider	Model ID	HF reference
SpeakLeash Foundation	`speakleash/Bielik-11B-v3.0-Instruct`	speakleash/Bielik-11B-v3.0-Instruct
OpenAI	`openai/gpt-oss-120b`	openai/gpt-oss-120b
Ministry of Digital Affairs of Poland	`CYFRAGOVPL/PLLuM-12B-instruct`	CYFRAGOVPL/PLLuM-12B-instruct
Meta	`meta-llama/Llama-3.1-8B-Instruct`	meta-llama/Llama-3.1-8B-Instruct
Meta	`meta-llama/Llama-3.3-70B-Instruct`	meta-llama/Llama-3.3-70B-Instruct
DeepSeek	`deepseek-ai/DeepSeek-R1-Distill-Llama-70B`	deepseek-ai/DeepSeek-R1-Distill-Llama-70B
MiniMaxAI	`MiniMaxAI/MiniMax-M2.5`	MiniMaxAI/MiniMax-M2.5
VillanovaAI	`VillanovaAI/Villanova-2B-2512-Preview-Apnea-FT`	VillanovaAI/Villanova-2B-2512-Preview-Apnea-FT
Mistral AI	`mistralai/Mistral-Small-4-119B-2603`	mistralai/Mistral-Small-4-119B-2603
Google	`google/gemma-4-31B-it`	google/gemma-4-31B-it

Large language models with vision capabilities

Vision-capable models can process text prompts together with supported image inputs. Use them only when your workflow requires visual understanding; for text-only tasks, select one of the large language models listed above.

Table 8 Large language models with vision capabilities
Provider	Model ID	HF reference
Mistral AI	`mistralai/Mistral-Small-4-119B-2603`	mistralai/Mistral-Small-4-119B-2603
Google	`google/gemma-4-31B-it`	google/gemma-4-31B-it

Embedding models

Embedding models convert text into vector representations. They are commonly used for semantic search, retrieval-augmented generation, document matching, clustering, and similarity checks. Use an embedding model when you need vectors rather than a generated text answer.

Table 9 Embedding models
Provider	Model ID	HF reference
Intfloat	`intfloat/e5-mistral-7b-instruct`	intfloat/e5-mistral-7b-instruct
BAAI	`BAAI/bge-multilingual-gemma2`	BAAI/bge-multilingual-gemma2
NovaSearch	`dunzhang/stella_en_1.5B_v5`	dunzhang/stella_en_1.5B_v5
sdadas	`sdadas/stella-pl-retrieval-8k`	sdadas/stella-pl-retrieval-8k

Choosing a model

For general text generation, start with an available large language model and change the model only when you need a different language profile, context behavior, latency, or output quality. For image-aware tasks, use a vision-capable model if one is available in your project. For search and retrieval workflows, use an embedding model and send the resulting vectors to your own vector database or retrieval layer.

If an API request fails, check whether the model ID was copied exactly, whether the model is available in your Sherlock AI project, and whether you are using the correct endpoint for chat completions or embeddings.

Still need help?

If this article doesn’t answer all of your questions, our support team will do their best to help you.

Contact support