Model Catalog
Browse model families, capabilities, and selection guidance for MX4 Platform deployments.
MX4 Platform supports multiple model families for chat, retrieval, and domain‑specific workflows. Availability and capacity depend on your deployment and licensing. Use Platform Studio to view the live catalog in your environment.
Catalog Overview
| Family | Use Cases | Notes |
|---|---|---|
| mx4-atlas-core | General chat, reasoning, tool orchestration | Primary routing target for mixed workloads |
| mx4-embed-v1 | Semantic search, retrieval, clustering | Arabic‑aware embeddings for RAG |
| Custom / Fine‑tuned | Domain‑specific and fine‑tuned models | Host your own checkpoints via Platform Serve |
Model IDs can differ by deployment. Use Platform Studio for the authoritative list.
Capabilities
Chat & Reasoning
Multi‑turn dialogue, instruction following, and agent‑style orchestration.
Embeddings
Arabic‑aware vector representations for search, clustering, and reranking.
Fine‑Tuning
Customize base models with your domain data and preferred tone.
Model Selection Guide
Choose Platform Core for
- Mixed workloads that need routing or tool usage
- Arabic‑first responses with strong general reasoning
- Enterprise workflows with strict sovereignty requirements
Choose Embeddings when
- You need semantic search across Arabic corpora
- You are building RAG pipelines or vector search
- You need deterministic, cacheable representations
Routing Profiles
Use routing profiles to balance quality, cost, and latency. Profiles are configured in Platform Studio and can be adjusted per request.
1{2 "model": "auto-router",3 "routing": {4 "profile": "balanced"5 }6}
Common profiles include balanced, quality, and cost. Availability depends on your deployment configuration.