Skip to main content

Model Support

LLMBoost provides validated support for a curated set of widely used production models. Each validated model is tested, optimized, and benchmarked on LLMBoost to ensure predictable latency, stable throughput, and efficient GPU utilization under production traffic conditions.

Validated Models

Our validated models include leading architectures such as Llama, Mixtral, Qwen, DeepSeek, and GPT-OSS families. These models have been evaluated on LLMBoost with demonstrated improvements in throughput and latency compared to baseline inference engines.

Loading supported models...


Need Support for a Specific Model?

Beyond the validated list, LLMBoost is designed to support a wide range of Hugging Face–compatible models. Our team can rapidly validate and optimize additional models based on your deployment requirements, including:

  • new open-weight model releases
  • custom fine-tuned models
  • workload-specific optimization needs

Contact us (contact@mangoboost.io) if you would like a specific model to be validated or optimized for your deployment.