Apertus 1.5
Switzerland's fully open 8B and 70B models, now with image and audio input, a 262K context and optional thinking, under Apache 2.0.
In short. Apertus 1.5 is a Swiss AI model that is open in the fullest sense: the weights, the data preparation steps, the training code and the evaluations are all public, and it was trained only on openly available data. It reads text, images and audio and answers in text, and it comes in a small 8B and a large 70B version.
Summary
ETH Zurich, EPFL and the Swiss National Supercomputing Centre (CSCS) released Apertus 1.5 on 24 July 2026. It continues the pre-training of the first-generation Apertus 1.0 models on a further 4 trillion tokens for the 8B model and 2 trillion for the 70B, for totals of 19T and 17T tokens. The architecture is unchanged: a decoder-only transformer with the xIELU activation, trained with the AdEMAMix optimizer. Images and audio enter as discrete tokens in the same sequence as text, through an Emu3.5-based image tokenizer and WavTokenizer, instead of through a separate vision encoder, and the models generate text only. The context window is 262,144 tokens, four times that of Apertus 1.0, and thinking is a flag of the chat template rather than a separate model.
Apertus is fully open: weights, intermediate checkpoints, data preparation pipelines, training code and evaluation harnesses are published under Apache 2.0, and the team says it trains only on openly available data, respects crawler opt-outs retroactively and excludes restrictively licensed material. In the technical report the 8B model leads fully open models of its class on knowledge, instruction following and cultural benchmarks, while specialized vision and audio models stay ahead on perception and on transcription of English and other high-resource languages. The project describes audio understanding as experimental. These are the developers' own evaluations.
The BF16 weights are about 18.4 GB for the 8B model and 144.6 GB for the 70B. At release the model card said support was not yet part of released Transformers or vLLM versions and pointed to Swiss AI's forks and a Docker image. The Hugging Face repositories are gated, so you have to accept the access terms before downloading. The card says no output filter is provided and recommends refreshing the weights about every six months to pick up data-deletion requests. Base and post-training stage checkpoints (SFT, DPO and RL) are published as separate repositories.
Good for
- Public bodies, universities and companies that need a model whose training data and process can be inspected and documented, for example for EU AI Act work
- Research on reproducible training, with base and stage-by-stage checkpoints published next to the final model
- Document understanding and long-form or lower-resource speech transcription, which the report names as the models' strongest multimodal areas
- Multilingual work, since the project targets broad coverage of languages and cultures
Watch out for
- Support was not yet in released Transformers or vLLM versions when the card was written, so you need Swiss AI's forks or its Docker image
- Tool calling is not supported in thinking mode, so pick one or the other per request
- Audio understanding is experimental, and the report says specialized models remain ahead on English and other high-resource speech
- The models read images and audio but generate text only, and video input is not supported, so extract frames and pass them as images
- The repositories are gated and there is no output filter, so add your own safeguards before putting it in front of users
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Estimates for the Apertus 1.5 8B variant. A mixture-of-experts model must hold all of its experts in memory even though only some are used for each token, so total parameters decide the memory and active parameters decide the speed.
| Variants | Parameters | BF16 | FP8 / INT8 | 4-bit (Q4) | Fits on |
|---|---|---|---|---|---|
| Apertus 1.5 8B BF16 | 8B | 18 GB | 9.2 GB | 5.52 GB | Laptop or phone |
| Apertus 1.5 70B BF16 | 70B | 161 GB | 81 GB | 48 GB | Workstation |
Estimates: weights plus about 15% overhead. Real usage grows with context length and depends on the software.
The license in plain words
One of the most widely used permissive licenses. You may use, change and sell products built on the model with only light conditions.
See also
- Apertus 1.5 vs OLMo 3.1: which fully open model should you use?
- Best open-source multimodal models in 2026 (images, video and audio)
- Open-weight AI models made in Europe
- Apertus 1.5: Switzerland's fully open model reads images and audio
- Open models that listen: nine take audio, but only three answer in speech
Frequently asked questions
- What is Apertus 1.5?
- Apertus 1.5 is a Swiss AI model that is open in the fullest sense: the weights, the data preparation steps, the training code and the evaluations are all public, and it was trained only on openly available data. It reads text, images and audio and answers in text, and it comes in a small 8B and a large 70B version.
- Can I use Apertus 1.5 commercially?
- Yes. Apertus 1.5 is released under Apache 2.0, a permissive license. You can use it commercially if you follow the standard conditions, such as keeping the license notice.
- Is Apertus 1.5 open source?
- Apertus 1.5 is released under Apache 2.0, a license approved by the Open Source Initiative. Training data and training code may not be public, so it is best described as open weights under an open-source license.
- Can I run Apertus 1.5 on my own computer?
- Yes. The smallest variant, Apertus 1.5 8B, needs about 5.52 GB of memory at 4-bit precision, which fits a modern laptop or a mid-range graphics card.
- When was Apertus 1.5 released?
- Apertus 1.5 was released by Swiss AI Initiative on July 24, 2026.
- How long a context window does Apertus 1.5 have?
- Apertus 1.5 supports a context window of up to 256K tokens.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- Apertus 1.5 (Apertus project, July 2026) apertus-ai.org
- Apertus 1.5: Building the next generation of open AI infrastructure (ETH AI Center, 24 Jul 2026) ai.ethz.ch
- Apertus v1.5 technical report (arXiv 2610.15542) arxiv.org
- Apertus-v1.5-70B model card (Hugging Face) huggingface.co
Release date and facts last verified November 11, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Olmo 3.1 Allen Institute for AI (Ai2) | 7.3B – 32.2B | 64K | Apache 2.0 (Permissive) | 5.04 GB Laptop or phone | |
| Qwen3.8-27B Alibaba Qwen | 27.8B | 256K | Apache 2.0 (Permissive) | 19 GB Gaming GPU | |
| Gemma 4 Google DeepMind | 5.1B – 30.7B | 256K | Apache 2.0 (Permissive) | 3.52 GB Laptop or phone | |
| Kolibri 1 Aleph Alpha | 78B 3.5B active | 1M | Apache 2.0 (Permissive) | 54 GB Workstation |