Apertus — local AI model by EPFL / ETH Zurich / CSCS
Written by Jakub Rusinowski · Last updated
Switzerland's open model, from EPFL, ETH Zurich and the Swiss National Supercomputing Centre. Trained on roughly 15 trillion tokens spanning more than 1,000 languages, with weights, data and process published under Apache-2.0 — the European sovereignty argument made concrete.
Variants
The smallest Apertus variant needs about 6 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Apertus 8B → 8B | ~5.6 GB | 65,536 | ollama run apertus | |
| Apertus 70B → 70B | ~43.1 GB | 65,536 | ollama run apertus |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Apertus locally
Install Ollama, then pull the tag.
ollama run apertusPick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Apertus 8B, Apertus 70BRecommended GPU
The cheapest catalogued GPU that runs Apertus locally (min 6 GB VRAM) is the Intel Arc B570 (10 GB).
Apertus — frequently asked questions
How much VRAM does Apertus need?
Apertus needs about 6 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Apertus 8B (6 GB, Q4_K_M); Apertus 70B (43 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Apertus on an RTX 4090 (24 GB)?
Yes — Apertus runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Apertus?
Q4_K_M is the best balance of quality and VRAM for Apertus in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Apertus with Ollama?
Apertus has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.