Vectara for generation
Vectara’s Mockingbird LLM leads the industry in accuracy for RAG workloads and structured outputs.
Choose Your Success: BYO LLM or RAG-optimized Generation
Vectara’s Mockingbird LLM reduces hallucinations and enhances structured output capabilities while maintaining low latency and cost efficiency. Vectara users can choose from a broad range of popular generative models to best design their workloads.
Vectara's new Mockingbird took HuckAI from being an overly polite librarian to giving answers I would expect from a senior coworker. The responses are clearer, easier to follow, and provide direct answers to difficult questions, helping our users get more work done. I switched immediately.
Bring-your-own LLM
Vectara doesn't dictate how you run your generation; we support multiple generative LLMs for agent success.
Vectara Mockingbird provides a RAG-optimized experience with industry-leading accuracy for enterprise workloads.
Vectara supports generative models from Open AI, Google, and many more.
Superior performance for RAG workloads
Mockingbird achieves the world's leading RAG output quality.
Its top-of-the-line hallucination mitigation capabilities make it perfect for enterprise RAG and autonomous agent use cases. Mockingbird outperforms GPT-4 and other major models on key metrics in RAG output and citation precision/recall.
Its specialized focus on RAG-specific tasks beats general-purpose models hands down.
Highest-precision structured outputs
Vectara Mockingbird generates structured outputs with levels of performance and accuracy that are ideal for running RAG workloads.
Structured output lets RAG connect to downstream tasks such as function calling or enabling agentic behavior. GPT-4 is the closest competitor to Mockingbird in precision.
It comes down to training the model on hard, real-world examples of structured data.
Embedded factual consistency score
Every Vectara response comes with an evolved version of Vectara’s popular open-source HHEM hallucination evaluation model.
HHEM detects the level of hallucinations in popular LLMs and in generated responses from those systems into the core platform.
The Factual Consistency Score on Vectara helps developers automatically assess hallucination. Users can use the new feature right out of the box to measure and improve response quality.
