Orchestration of Experts
The Leeroo Orchestrator harnesses expert LLMs to establish the next level of LLMs. Through a self-play learning loop, it not only optimizes expert selection but also guides the development of highly specialized expert models for enhanced performance. Experience AI advancement with the Leeroo Orchestrator, where specialized expertise meets continuous improvement.
Orchestrator harnesses the full potential of state-of-the-art LLMs today, setting a new benchmark in AI efficiency and effectiveness.
Thanks to its self-play learning loop, the Orchestrator continually refines its capabilities, ensuring each iteration surpasses the last in understanding and application.
Strategic Development Guidance
It doesn't just orchestrate; it informs. By identifying where the gaps lie in the AI landscape, the Orchestrator directs us to develop the next wave of expert models, maximizing technological impact.
Dynamic Learning and Adaptation
With each new expert model introduced, the Orchestrator seamlessly integrates and learns to utilize these resources, becoming ever more specialized and adaptable.
Leading the Way:
Top Open-Source LLM in MMLU Benchmark
Leveraging small open-source LLMs
Top Open-Source LLM: Achieves 76% accuracy on the MMLU benchmark, outperforming Mixtral at equal cost.
Cost-Efficient: Matches GPT3.5's performance at nearly 1/4th the expense.
Broad Compatibility: Runs on consumer GPU and is deployable across cloud services or on-premises.
For more details, click here.
New architecture, state-of-the-art results
Innovative Architecture: Transcends traditional MoE models, enabling each expert to operate independently across diverse platforms and architectures.
Matching GPT4 Performance: Achieves parity with GPT4 on key benchmarks at 50% of the cost.
Surpassing GPT4: Exceeds GPT4's capabilities with 75% of the cost, setting a new standard for efficiency and effectiveness.