Swiss fully-open model Apertus ships 1.5 with multimodal input and reasoning
sull · x · 2026-11-22
Apertus, a fully open foundation model from the Swiss AI Initiative (EPFL, ETH Zurich, CSCS), promotes open weights, open data, and open science — training data, code, weights, methods, and alignment principles are all documented and reproducible.
- 8B and 70B scales, claimed to be competitive with top open models at equivalent size
- Trained from day one on 1000+ languages
- Techniques for honoring opt-outs, removing PII, and preventing memorization
- Oct 14: Apertus 1.5 released, adding multimodal input, reasoning, longer context, and better instruction following
- Swisscom is a strategic partner of the initiative
More from Models
- Partial model collapse: LLMs quietly degrade on fuzzy tasks while excelling at verifiable ones — maksym_andr · 2026-11-22
- Theo's new video argues open models are doomed: "If you like Open Models, I'm sorry" — gnukeith · 2026-11-22
- GPT-6.2 breaks Responses API prompt caching: hit rate falls from 97.4% to 46.4% — lxfater · 2026-11-22
- Reddit user claims Gemini knew his keyboard's USB device name, then denied it — Pyro2745 · 2026-11-22
- Cryptographer warns frontier AI math is already good enough to break cryptography — rickasaurus · 2026-11-22
- Tesla FSD V15 Brakes Early for Pedestrians in Heavy Fog, Driver Calls It Life-Saving — yunta_tsai · 2026-11-22