Knowledge Hub · Sovereign AI

Which AI models are available on FirstNet's AI Factory by FirstCoreAI?

AI Factory · Answered by FirstNet Technology Services

Short answer

Eight production endpoints are live on the FirstCoreAI AI Factory, all open-weight models running in FirstNet's Durban data centre.

In detail: Available AI models on the AI Factory by FirstCoreAI

The catalogue, verified on 18 September 2026:

  • Gemma 4 31B: general chat and agents, 128K context and tool calling
  • Gemma 4 E4B: lightweight chat, 32K context, tools and basic vision
  • Qwen3.8 27B: coding, agents and vision, 128K context
  • Qwen3-VL 8B: document and image understanding, 128K context
  • Whisper Large v3: speech to text
  • MOSS-Transcribe-Diarize: transcription with speaker separation, up to 10 minutes per request
  • BGE-M3: multilingual embeddings, 1024 dimensions and 8K context
  • BGE reranker v2 M3: retrieval reranking for search and RAG

Every model is served through an OpenAI-compatible API (chat completions, embeddings, audio transcriptions and rerank), and an Anthropic-compatible API is also available. FirstCoreAI, FirstNet's AI business unit, benchmarks each model on its own hardware against frontier APIs for accuracy, latency, throughput under real concurrency and cost per unit of work before it goes into production.

Where a workload genuinely needs an overseas frontier model, requests can be routed to one, and FirstCoreAI tells you when a request leaves the country. To see the models running on your own workload, book a demo through the form on FirstNet's AI Factory page.

Source: FirstNet AI Factory service page →

Call