Access to o1 preview and o1 mini in the EU

At Saga, as an LLM-agnostic platform, we are committed to ensuring that our clients have access to the newest models that meet their specific needs, while maintaining the highest security and privacy standards. We are excited to announce that we now have access, as one of the first in the EU, to OpenAI’s o1 preview and o1 mini running in the EU. For our clients in Europe, the o1 models are fully compliant with GDPR and other local data protection regulations. This ensures that the adoption of this technology aligns with the high standards of data security and confidentiality expected in the legal sector.

On Vals.ai's Legalbench, where they evaluate LLMs on a wide range of open source legal reasoning tasks, o1 Preview currently holds the highest score. However, it is also the most expensive and slowest model. Deepseek’s entrance will most likely result in a decrease of this price very soon.

As you can see, AI is not perfect, always check the output.

What are reasoning models and what is o1?

The o1 models represent a significant step forward in AI, marking the start of the reasoning models. Reasoning models like o1 are designed to perform complex reasoning tasks. These models are trained to think through problems systematically before providing answers, similar to human thought processes. Reasoning models rely mostly on Reinforcement Learning, very much like the AI models that are so good at Chess and the Chinese game Go.

What can reasoning models do for legal professionals?

Reasoning models like o1 offer significant potential for legal professionals, enhancing their capabilities and efficiency in various aspects of legal work:

• Legal reasoning: These models can enhance case-based reasoning by understanding the context of a case better and evaluate the validity of legal arguments.
• Risk modeling and scenario analysis: Legal teams can use o1 to model hypothetical scenarios, stress-test agreements, and look at things from different perspectives.
• Risk detection: reasoning models are better at identifying minor yet potentially consequential terms and errors in legal briefs, offering a more refined understanding of nuanced legal distinctions.
• Legal research and information processing: These models can efficiently process vast amounts of legal information, aiding in comprehensive legal research and analysis.

Adoption remains key

Our adoption program ensures that users at every level of experience can maximize the benefits of the o1 models. From introductory workshops to advanced training, we guide legal professionals through each stage of the process, ensuring that they feel confident and empowered to use the platform effectively.