Skip to main content

Ministral 3 14B Instruct

Mistral's 14B instruct model — multimodal (text + image), served on Intel Arc.

Overview​

  • model_name: hf-mistralai-Ministral-3-14B-Instruct-2512 (HuggingFace import, currently free)
  • Source: mistralai/Ministral-3-14B-Instruct-2512
  • Scale: 14B parameters, Mistral3 architecture with the Pixtral vision tower
  • Served on: GPU via vLLM, tensor-parallel across 2× Intel Arc Pro B70 (32GB) — vision enabled (ColabHive ships two XPU fixes for the upstream Pixtral path)
  • Status: ready (verified end-to-end through the public inference route, 2026-08-03)

When to use​

✅ General instruction-following and chat with strong multilingual quality, plus image understanding (describe, extract, reason over images) in the same endpoint.

❌ For pure code work use Qwen3 Coder 30B; for heavy reasoning chains use DeepSeek R1-0528.

Live specs​

curl "https://api.colabhive.com/api/builder/v1/endpoints?visibility=public&search=Ministral-3"

Quick start​

curl -X POST "https://api.colabhive.com/api/builder/v1/endpoints/<endpoint_id>/infer" \
  -H "X-API-Key: $COLABHIVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": {"messages": [{"role": "user", "content": "Summarize the key differences between TCP and UDP."}], "max_tokens": 400}}'