Apertus-v1.5-8B-SFT

Apertus 1.5 is a family of 8B and 70B fully open, multilingual, and multimodal models from the Swiss AI Initiative, trained only on openly available data and released with open weights, open data, and open training documentation. The models cover a large variety of languages, handle contexts up to 262,144 tokens, accept image and audio input, and support a thinking mode for reasoning tasks.

This checkpoint is the supervised finetuning stage of the 8B post-training pipeline. All nine checkpoints of the pipeline are released.

Evaluation

The leftmost columns are the Apertus 1.5 8B lineage, each column adding one stage to the one before it. The final row averages all listed benchmarks.

Thinking disabled

Thinking enabled

Long context

How to Use

Users of Apertus can find instructions on getting started with desktop software and cloud providers on our website. Please visit the documentation page if you are interested in trying the model.

Deployment of the models is supported in the following open source frameworks: Transformers, vLLM.

vLLM

We are currently working on adding support for our models to upstream vLLM and transformers releases. In the meantime, you can use our modified version of vLLM and transformers to run the models. We have pre-installed the dependencies in a Docker image which is available in the GitHub Container Registry. The source dockerfile used to build the image is available in this repository. You can pull the image with the following command:

• amd64 architecture:

  docker pull ghcr.io/swiss-ai/vllm_apertus_1.5_release:latest-amd64
  

• arm64 architecture:

  docker pull ghcr.io/swiss-ai/vllm_apertus_1.5_release:latest-arm64
  

You can use the following command to run the model with vLLM:

vllm serve swiss-ai/Apertus-v1.5-8B-SFT \
  --chat-template-content-format string \
  --gpu-memory-utilization 0.6 \
  --max-model-len 262144 \
  --enable-auto-tool-choice \
  --tool-call-parser apertus

Depending on your hardware, you may need to adjust --tensor-parallel-size, --gpu-memory-utilization, and --max-model-len (e.g. lower --max-model-len if you run out of memory). On some hardware configurations, CUDA Graph capture may fail with --tensor-parallel-size > 1 due to the fused all-reduce RMS optimization. If this occurs, launch vLLM with --compilation-config.pass_config.fuse_allreduce_rms false.

Instructions for Launching Apertus 1.5 with Thinking Mode Enabled

To enable thinking mode, set --reasoning-parser and --default-chat-template-kwargs.enable_thinking as shown below. The tool-call flags are intentionally omitted: tool calling is unsupported in thinking mode, so we don't recommend combining the two.

vllm serve swiss-ai/Apertus-v1.5-8B-SFT \
  --served-model-name swiss-ai/Apertus-v1.5-8B-SFT-thinking \
  --chat-template-content-format string \
  --gpu-memory-utilization 0.6 \
  --max-model-len 262144 \
  --reasoning-parser apertus \
  --default-chat-template-kwargs.enable_thinking true

Transformers

Apertus 1.5 accepts interleaved text, image, and audio inputs and generates text. The model does not generate audio or images.

The integration is not yet part of a released Transformers version (upstreaming is in progress). Until then, install transformers from our branch:

pip install "transformers[torch,vision,audio] @ git+https://github.com/swiss-ai/transformers.git@3797303dda74844e3d1f8977ff5518bb91f818b4"

Load the processor and model once:

import torch
from transformers import AutoModelForMultimodalLM, AutoProcessor

MODEL_ID = "swiss-ai/Apertus-v1.5-8B-SFT"

processor = AutoProcessor.from_pretrained(MODEL_ID)
model = AutoModelForMultimodalLM.from_pretrained(
    MODEL_ID, dtype="auto", device_map="auto"
).eval()


def generate(messages, max_new_tokens=256, **template_kwargs):
    inputs = processor.apply_chat_template(
        messages,
        add_generation_prompt=True,
        tokenize=True,
        return_dict=True,
        return_tensors="pt",
        **template_kwargs,
    ).to(model.device)
    with torch.inference_mode():
        output_ids = model.generate(**inputs, max_new_tokens=max_new_tokens)
    return processor.decode(
        output_ids[0, inputs["input_ids"].shape[-1]:], skip_special_tokens=True
    )

Text

messages = [
    {"role": "system", "content": "You are a concise and helpful assistant."},
    {"role": "user", "content": "Explain why the sky appears blue in one sentence."},
]

print(generate(messages))

Image

Images can be supplied as URLs, local paths, PIL images, or arrays:

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/coco_sample.png",
            },
            {"type": "text", "text": "Describe this image in detail."},
        ],
    }
]

print(generate(messages))

Audio

Audio files referenced by URL or local path are decoded and resampled to 24 kHz automatically; in-memory waveforms are accepted as mono NumPy arrays already sampled at 24 kHz:

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Summarize what is said in this audio clip."},
            {
                "type": "audio",
                "url": "https://huggingface.co/datasets/hf-internal-testing/dummy-audio-samples/resolve/main/bcn_weather.mp3",
            },
        ],
    }
]

print(generate(messages))

Batching

A batch may mix prompts with different numbers and kinds of media. The shipped tokenizer defaults to the left padding that batched generation requires:

conversations = [
    [
        {
            "role": "user",
            "content": [
                {
                    "type": "image",
                    "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg",
                },
                {"type": "text", "text": "Describe this image in one sentence."},
            ],
        }
    ],
    [{"role": "user", "content": "Name the four official languages of Switzerland."}],
]

inputs = processor.apply_chat_template(
    conversations,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
    processor_kwargs={"padding": True},
).to(model.device)

with torch.inference_mode():
    output_ids = model.generate(**inputs, max_new_tokens=256)

print(processor.batch_decode(output_ids[:, inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

Thinking mode

Thinking is turned off by default. Pass enable_thinking=True to apply_chat_template to activate it; the model then reasons between the <|inner_prefix|> and <|inner_suffix|> tokens before its visible answer.

messages = [
    {"role": "user", "content": "A bat and a ball cost 1.10 CHF together. The bat costs 1 CHF more than the ball. What does the ball cost?"},
]

print(generate(messages, enable_thinking=True, max_new_tokens=2048))

Budget generously for max_new_tokens in thinking mode: the reasoning can be several times longer than the visible answer, and a truncated generation may end before the answer begins. The reasoning markers are deliberately not stripped by skip_special_tokens=True, so the deliberation span can be parsed out of the decoded text.

Input notes

• Each image or audio content block corresponds to exactly one media item; placeholder/media count mismatches raise an error instead of being silently reassigned.
• Audio contributes 40 tokens per second.
• The LM head covers the 131,072 text tokens (output_vocab_size); logits are padded to the full 266,752-token vocabulary with non-selectable scores, so standard generation utilities work unchanged while image and audio tokens are never generated.
• The vision and audio tokenizers are precision-sensitive and stay in float32 automatically on half-precision loads. Avoid re-casting the loaded model with .half()/.to(dtype).
• Native video inputs are not supported; applications may extract frames and pass them as images.

Post-Training Pipeline

Each stage starts from the checkpoint produced by the stage before it.

1. Supervised finetuning. Trains on a mixture of text and multimodal instruction data. Chain-of-thought traces are attached only to the prompts that a non-reasoning model fails under pass@k. A dedicated long-context stage at the 262,144-token window closes the supervised phase.
2. Reinforcement learning. Optimizes verifiable rewards, computed by checking or executing the model's output: mathematics is verified against ground truth, instruction-following prompts run formal constraint checkers, and code runs against test cases. The model emits final answers through a dedicated answer tool. Training mixes thinking-enabled and thinking-disabled prompts.
3. Preference alignment. Direct Preference Optimization. Offline DPO trains on MaxMin_Tr_3600-Filtered-Decontaminated; online DPO uses the prompts of that dataset. The 8B model trains with online DPO; the 70B model trains with offline DPO followed by online DPO. Both forms run with thinking disabled.
4. Charter alignment. A short self-distillation stage (SDPO). The student answers a prompt, a judge writes textual feedback on how well the answer follows the Apertus charter, and the teacher signal is obtained by conditioning the student on that feedback. Applied to the 8B model.

Data mixtures, hyperparameters, and ablations are documented in the technical report.

This checkpoint

The supervised-finetuning checkpoint, after the dedicated long-context stage. It has undergone no reinforcement learning and no preference alignment.

Limitations

Apertus can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.

See the Apertus Charter for the alignment principles that govern the responses of the models.

License

Apache 2.0.

Legal and Compliance

By using the Apertus LLM you agree to indemnify, defend, and hold harmless ETH Zurich and EPFL against any third-party claims arising from your use of Apertus LLM.

EU AI Act documentation (Apertus_1_5_EU_Public_Summary.pdf and Apertus_1_5_EU_Code_of_Practice.pdf) is available at https://github.com/swiss-ai/apertus-legal/, alongside the usage policy and privacy policy.

We strongly advise downloading and updating new model weights from this site every six months.

Data protection contacts: llm-privacy-requests@swiss-ai.org and llm-copyright-requests@swiss-ai.org.

Contact

General inquiries: llm-requests@swiss-ai.org, or https://apertus-ai.org/contact/.

Citation

@misc{apertus2026apertus15opentext,
      title={Apertus 1.5: From Open Text Models to Open Multimodal Reasoners}, 
      author={Project Apertus and : and Yixuan Xu and Yassine Turki and Vinko Sabolčec and Valentina Pyatkin and Tommy Chu and Sven Najem-Meyer and Sean F. Carroll and Robert Smith and Raphael Kreft and Oscar Alessandro Tazza and Oleg Lavrovsky and Nicola Irmiger and Matteo Santelmo and Marian Schneider and Juan García Giraldo and Hanna Yukhymenko and Eduard Ďurech and Davit Melikidze and Dani Tamayo and Bettina Messmer and Aryan Ahadinia and Anunay Yadav and Allen Huang and Ahmad Fraij and Alessandro Bombarda and Alex Padula and Alexander Hägele and Alexander Sternfeld and Anna Hedström and Anna Kosovskaia and Arthur Wuhrmann and Arundhati Balasubramaniam and Badr Al Mahouri and Barna Pasztor and Bartosz Szostakiewicz and Bich Ngoc Doan and Bryan Schmid and Burak Sezen and Camille Challier and Clément Charmillot and Dana Shayakhmetova and Daniel Perruchoud and David Brüggemann and David Isztl and Elia Palme and Emma Anastassova and Fabian Bösch and Faruk Zahiragić and Fawzi Roberto Mohamed and Fiona Panicali and Florian Kolly and Flynn Guerra and Francesco Monti and Frederike Lübeck and Gennaro Oliva and Haofeng Yang and Haoliang Shang and Helen El Khoury and Henrique Mendonça and Jannis Vamvas and Jérome Landolt and Jessica Lam and Jingchun Xiao and Jonathan Coles and Joshua Oyebanji and Khanh Nguyen and Konstantinos Pitas and Lasse B. Strand and Li Mi and Loïc Deslarzes and Lorenzo Baggi and Lorenzo Noci and Luca Baumann and Luca Sartori and Lukas Drescher and Marco De Negri and Marco Passerini and Mark Klein and Marko Stojanovic and Martin Majsec and Matin Ansaripour and Mattea Busato and Mauro Pellonara and Max Neuwinger and Michael Graber and Michal Tešnar and Michelle Wastl and Miguel Gila and Mikolaj Boronski and Morgane Magnin and Moritz Reihs and Mukhammadali Sayfiddinov and Nathan Ranchin and Nicolas Cuveillier and Oscar Wohlfahrt and Qianqian Zou and Rada Kamysheva and Rita Sevastjanova and Roman Machacek and Sebastian Keller and Simon Matrenok and Simon Pierre and Stefano Schuppli and Suei-Wen Chen and Szabina Horváth-Mikulas and Taliesin Perez and Thilo Spinner and Thomas Kleine Buening and Tim Robinson and Tymoteusz Kwiecinski and Vilém Zouhar and Vinzenz Thoma and Xiaozhe Yao and Yannick Metz and Yuhei Fukuhara and Yusif Askari and Zixuan Chen and Alexander Ilic and Ana Klimovic and Andreas Krause and Andrei Kucharavy and Elliott Ash and Joost VandeVondele and Martin Rajman and Mennatallah El-Assady and Thomas Schulthess and Timo Kehrer and Torsten Hoefler and Antoine Bosselut and Martin Jaggi and Imanol Schlag},
      year={2026},
      eprint={2610.15542},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2610.15542}, 
}