New Metamodel Uses Level 2 AI Reasoning Capabilities to Surpass GPT-4o by 5 Points on AlpacaEval 2.0, Opening a New Era in Performance While Radically Reducing Costs
NEW YORK, NY / ACCESSWIRE / September 4, 2024 / Touchcast today unveiled the TM1, a Level 2 reasoning AI metamodel powered by Agentic Pipelines as a Service (APAS). TM1 represents a significant advancement in AI capabilities, delivering unprecedented performance through a single, elegant API call.
TM1 has achieved a score of 62 on the AlpacaEval 2.0 benchmark, surpassing GPT-4o and establishing itself as the most accurate AI metamodel available. This advancement in accuracy is complemented by the lowest inference cost on the market, positioning TM1 as the definitive solution for advanced AI workloads.
Beyond Level 1: A New Era of AI Intelligence
While LLMs have revolutionized many aspects of AI, they often fall short when faced with tasks requiring nuanced reasoning, multi-step problem-solving or domain-specific expertise. Agentic Pipelines as a Service (APAS) represents a significant evolution beyond traditional LLMs by orchestrating multiple specialized agents leveraging multiple LLMs, each contributing its unique strengths to the task at hand.
APAS enhances inference by adding the following capabilities:
Reflective reasoning: The ability to analyze and refine its own outputs
Deliberate processing: Taking time to consider and improve responses
Meta-cognitive abilities: Assessing the quality of its own outputs
Multi-step problem-solving: Breaking down complex tasks into manageable subtasks
These capabilities allow TM1 to produce more thoughtful, accurate, and nuanced responses compared to traditional Level 1 models.
"Most of today's Large Language Models operate at what is defined as Level 1 intelligence," said Edo Segal, Touchcast founder and CEO. "These models generate quick responses but lack self-correction mechanisms and metacognitive abilities. TM1 introduces Level 2 AI. By leveraging our Mixture of Agents approach and our recent integration with Groq for hyper-fast inference speeds, we're providing a new level of inference output that is both incredibly powerful and incredibly cost-effective."
Key features and benefits of the Touchcast TM1 model include:
Unparalleled Accuracy: TM1 outperforms all other models on the comprehensive AlpacaEval 2.0 benchmark, setting a new industry standard.
Radical Cost-Efficiency: At $0.15 per million tokens for input and $0.50 per million tokens for output, TM1 is 33x more cost-effective than leading competitors, offering substantial savings without compromising on quality.
Simplified Agentic AI: APAS distills complex AI processes into a single, powerful LLM call, making advanced AI capabilities easily accessible without the need for complex infrastructure.
Industrial-Grade Scalability: Built on the Touchcast Platform, TM1 is engineered for enterprise-level deployment, eliminating the need for in-house agentic pipeline development and maintenance.
The Touchcast TM1 model is now available in preview. For more information, visit https://touchcast.com/tm1.
About Touchcast
Touchcast is revolutionizing AI infrastructure with the world's fastest and most cost-effective inference ecosystem, providing unparalleled access to leading AI models, redefining AI performance and efficiency.
For more information, visit www.touchcast.com.
Contact Information
Touchcast Media Desk
press@touchcast.com
SOURCE: Touchcast
View the original press release on newswire.com.
