Key facts

Founded
2019
HQ
San Francisco, United States
Employees
—
Funding stage
Private · Qualcomm-owned
Further facts
  • Ownership: BentoML joined Modular on 10 February 2026 as a strategic product acquisition, and Qualcomm completed its acquisition of Modular on 29 July 2026.
  • Timeline: The open-source BentoML project continues under the Apache 2.0 licence, with the same docs, community and contribution process.
  • Leadership: Chaoyu Yang is the founder of BentoML, and Tim Davis is senior vice-president and general manager of Modular at Qualcomm.

Focus areas

LLM inference Open source GPU cloud On-premises deployment

About

BentoML was started in 2019 to give machine learning teams a simpler path from model to production. The open-source BentoML framework packages models of any architecture, framework or modality with their dependencies, and the bentoml.com inference platform adds deployment automation, scaling, observability and access control across clouds and GPU hardware. More than 10,000 organisations, including more than 50 Fortune 500 companies, use BentoML to deploy models. On 10 February 2026 BentoML joined Modular as a strategic product acquisition, and Qualcomm completed its purchase of Modular on 29 July 2026; the open-source project continues under the Apache 2.0 licence.

How they use AI

BentoML is infrastructure for running AI models rather than a model maker. Teams define a model service in Python, set its GPU resources, and deploy open-source or fine-tuned models such as Llama through one framework. The platform handles distributed LLM inference across multiple GPUs, scaling tuned to inference metrics, performance tuning against latency, throughput or cost targets, and day-one access to newly released open-source models. Inside Modular, BentoML is paired with Modular's MAX serving stack so the same deployments run across NVIDIA, AMD and future accelerators.

Products and services

  • BentoML open source — Apache 2.0 framework for packaging and serving models of any architecture, framework or modality
  • Bento inference platform — Managed platform for deploying, scaling, monitoring and optimising model inference
  • Open Source Model Launcher — Pre-optimised open-source models ready to deploy, with day-one access to new releases
  • Compute orchestration — Multi-cloud GPU orchestration with resource and quota tracking and fine-grained access control

Notable customers / partners

  • Neurolabs

Key people

  • Chaoyu Yang — Founder, BentoML