Just launchedThe AI FactoryA sovereign AI Factory for South African businesses AI Factory by FirstCoreAI delivers GPU and AI inference as a service: private inference endpoints, dedicated or fractional GPU capacity and a curated catalogue of eight open-weight models, all running in South Africa and reachable through familiar OpenAI-style APIs.
Watch: Local is lekker
Local is lekker, and so is local AI. In 36 seconds: open-weight models on our own GPUs in South Africa, with your data kept in the country.
- Open-weight models on FirstNet GPUs, in South Africa
- Your data never leaves the country
- Never used to train models
- Billed in rand, no exchange rate fluctuations
Read the film as text
- Local is lekker. And when it comes to AI, local is better. FirstNet's AI Factory runs open-weight models on our own GPUs, right here in South Africa. Your data never leaves the country, and it's never used to train models. Billed in South African currency. No exchange rate fluctuations. Book a demo at firstnet.co.za.
The FirstNet AI Factory
Watch on its own page →Read the transcript
Local is lekker. And when it comes to AI, local is better. FirstNet's AI Factory runs open-weight models on our own GPUs, right here in South Africa. Your data never leaves the country, and it's never used to train models. Billed in South African currency. No exchange rate fluctuations. Book a demo at firstnet.co.za.
The AI Factory service
- Private inference endpoints hosted locally; nothing leaves the country
- Eight live production model endpoints on the AI Factory
- Dedicated or fractional GPU reservation, sized to your workload
- Optional frontier routing, with notice when a request goes offshore
- Rand-denominated, forecastable commercials with no FX surprises
Hardware and models
The AI Factory is physical compute that FirstCoreAI, FirstNet's AI business unit, operates in FirstNet's South African data centre, tuned for production throughput and concurrency rather than benchmark leaderboards.
Compute
Current-generation NVIDIA Blackwell-class GPUs with AMD EPYC processors and more than a terabyte of system RAM per node.
Models
Eight endpoints, catalogue verified 18 September 2026: Gemma 4 31B for general chat and agents (128K context), Gemma 4 E4B for lightweight chat, Qwen3.8 27B for coding and vision (128K context), Qwen3-VL 8B for documents and images, Whisper Large v3 for speech to text, MOSS-Transcribe-Diarize for transcription with speaker separation, BGE-M3 multilingual embeddings and the BGE v2 M3 retrieval reranker.
Storage and fabric
A multi-terabyte NVMe model cache shared over NFS, a high-throughput in-band fabric between nodes and segregated out-of-band management.
Software stack
Kubernetes with Run:ai for GPU-aware scheduling and vLLM for inference, with FirstCoreAI's own API proxy in front handling keys, metering, spend caps and per-key limits, plus Grafana and Prometheus for operations.
Drop-in OpenAI compatibility
You get an API key and point any OpenAI SDK at https://api.firstcore.ai/v1, with no other code changes. An Anthropic-compatible API is also available. Every key carries its own limits and spend cap, you see usage and spend per key as it happens, and traffic is encrypted with TLS throughout. FirstCoreAI sees usage and spend, never your prompts or outputs.
- /v1/chat/completions for chat and coding
- /v1/embeddings for search and retrieval
- /v1/audio/transcriptions for speech-to-text
- /v1/rerank for retrieval reranking
The details
Performance on the production clusterFirstCoreAI stress-tested the live cluster under real serving conditions.
Performance on the production cluster
FirstCoreAI stress-tested the live cluster under real serving conditions. Every model is also benchmarked on the same hardware against frontier APIs for accuracy, latency, concurrency and cost per unit of work, so you see the evidence before committing.
~0.6s
Median time to first token on chat and coding models.
271
Concurrent requests sustained at peak during the full stress run.
0
Errors across all stress tests, at every concurrency level.
Is the AI factory the right fit?The AI Factory suits high-volume document and call processing, RAG and retrieval back-ends, internal…
Is the AI Factory the right fit?
The AI Factory suits high-volume document and call processing, RAG and retrieval back-ends, internal assistants, classification and extraction, and any workload where data sensitivity or cost rules out a foreign API. To see your workload on the Factory, book a demo or a tour below.
- Dedicated capacity instead of a shared public queue
- Every inference stays inside the South African perimeter
- Support from the engineers who run the Factory
Where is the AI Factory hosted?
In FirstNet's data centre in South Africa. FirstCoreAI owns and operates the GPU cluster, and inference happens inside South Africa.
Do I need to rewrite my application to use the AI Factory?
No. Every model is exposed as an OpenAI-compatible API, so any SDK that already works with OpenAI works with the AI Factory. You only swap in the endpoint URLs and credentials issued at onboarding.
Which models are available?
Eight production endpoints are live: two general chat models, a coding and vision model, a documents and images model, speech to text, transcription with speaker separation, multilingual embeddings and a retrieval reranker. Frontier APIs can be routed to where a workload needs one.
How is access secured?
Clients authenticate with OAuth2 client credentials and receive short-lived bearer tokens. TLS is terminated at the cluster edge and OAuth2 proxies validate each request.
Can I get dedicated GPU capacity?
Yes. The AI Factory offers both dedicated and shared GPU options, sized to the workload, with rand-denominated billing.
Can I tour the AI Factory?
Yes. FirstCoreAI runs 60 to 90 minute tours covering the hardware, the model catalogue and a live workload. Book one with the form on this page.
From the Knowledge Hub
In-depth answers about AI Factory
See it running on home soil
Where sovereign AI fits
- Sovereign AI AI that stays home, in every layer.
- Voice Every call, clear.
- Security Threats, contained.
- Cloud Data, protected.
- Connectivity Always connected.
The intelligence layer sits on top of everything FirstNet runs, and can work with data from every layer below it. See the five layers
Book an AI Factory demo
Bring one workload that should not leave the country. We show it running sovereign, and say so if it would be better off elsewhere.
- Tell us the workloadA document pipeline, call archive, chatbot or a model you cannot deploy.
- We scope it with youA short call with the engineers who run the AI Factory.
- See it runningA live demo, or a 60 to 90 minute tour of the AI Factory.