Nexusflow’s NexusRaven-V2: Revolutionizing Open-Source Generative AI for Copilots and Agents

TL;DR:

  • Nexusflow introduces NexusRaven-V2, a 13-billion parameter open-source generative AI model.
  • NexusRaven-V2 excels in function calling, translating human instructions into precise software tool actions.
  • Outperforms GPT-4 with a 7% higher success rate in tool utilization.
  • The data curation process combines open datasets and Meta’s Llama ecosystem.
  • NexusRaven-V2 offers enterprise clients full ownership and customization of the AI model.
  • Potential to revolutionize workflow automation with a smaller model size and higher accuracy.
  • Nexusflow commits to superior AI quality, reliability, cost reduction, and safety.
  • The company opens its model artifacts to the community to advance enterprise generative AI adoption.

Main AI News:

In a bold move that is set to redefine the landscape of generative AI, Nexusflow, a pioneering force in the cybersecurity sector, has officially introduced NexusRaven-V2. This remarkable open-source generative AI model boasts a staggering 13 billion parameters and offers a game-changing function-calling capability. In essence, NexusRaven-V2 possesses the unique ability to comprehend human instructions and seamlessly translate them into precise function/API calls, thereby unleashing a vast array of software tools. At the heart of the OpenAI Assistants API, this function-calling prowess holds the key to empowering copilots and agents with unparalleled software tool utilization.

Derived from the highly acclaimed CodeLlama-13B model, NexusRaven-V2 has set a new benchmark by achieving an impressive 7% higher success rate in tool utilization compared to the latest iteration of OpenAI’s GPT-4 model, as demonstrated on meticulously curated general software tool use benchmarks.

Nexusflow’s approach to data curation is nothing short of revolutionary, as it leverages open datasets and harnesses the formidable capabilities of Meta’s Llama ecosystem to curate an extensive repository of high-quality data. This data serves as the bedrock for training NexusRaven-V2. Notably, the data curation and training processes of NexusRaven-V2 do not rely on proprietary LLMs like OpenAI’s GPT-4. This groundbreaking approach empowers enterprise clients to have full ownership of the model used to construct copilots and agents. Such autonomy is particularly crucial for enterprise applications that demand real-time information, unwavering quality, enhanced safety, and profound domain-specific customization.

With its unparalleled quality and efficiency, NexusRaven-V2 has the potential to trigger a paradigm shift in the realm of workflow automation for complex software. It achieves this with a significantly more compact model size and an elevated level of accuracy.

Jiantao Jiao, the CEO and co-founder of Nexusflow, proudly declared, “NexusRaven-V2 outperforms OpenAI’s GPT-4 model head-to-head when it comes to harnessing the power of software tools. This accomplishment underscores Nexusflow’s technical prowess in delivering enterprise solutions built on open-source models.”

NexusRaven-V2 is a testament to Nexusflow’s unwavering commitment to creating effective, dependable generative AI solutions that grant customers absolute control over their models and data. Key advantages for Nexusflow’s clients include:

  1. Ownership of cutting-edge generative AI models without the data risks associated with proprietary vendors.
  2. Enhanced generative AI quality, reliability, cost-effectiveness, and safety, complete with up-to-date customization. This stands in stark contrast to solutions reliant on proprietary model vendor APIs, which often suffer from unpredictable quality degradation and lag behind in knowledge.

Nexusflow remains steadfast in its pursuit of thought leadership in the enterprise generative AI arena. To further accelerate the adoption of generative AI using open-source models, the company is generously opening its model artifacts to the broader community, fostering collaboration and innovation on an unprecedented scale.

Conclusion:

Nexusflow’s NexusRaven-V2 represents a significant leap in open-source generative AI technology, with the potential to disrupt the market by offering unparalleled control and customization to enterprise clients. This innovation not only improves workflow automation but also ensures data security and reliability, positioning Nexusflow as a key player in the evolving landscape of AI solutions.

Source