Modal Labs

What Modal Labs does

Modal Labs operates a serverless GPU cloud platform purpose-built for AI workloads. The platform provides an AI-native runtime with sub-second container cold starts, instant autoscaling from zero to over 1,000 GPUs, and multi-cloud GPU routing across regions in real time. The product portfolio rests on a Core Platform (the underlying container runtime) and includes three core products — Inference (LLM and multi-modal inference serving with vLLM and OpenAI-compatible APIs), Training (single-GPU fine-tuning up to 128x B200 multi-node training with 3,200 Gbps Infiniband, RL, and parallel hyperparameter sweeps), and Sandboxes (isolated, ephemeral GPU environments for coding agents, background agents, and RL rollouts at hundreds of thousands of concurrent runs) — plus Batch and Notebooks modules. The entire platform is programmable from Python via the Modal SDK, with no YAML or infrastructure configuration files required.

The business model is usage-based, charging per second of GPU compute, with a $30 free credit for new sign-ups and a fully self-service product-led go-to-market. Customers span AI-native startups (Suno, Runway, Decagon, Physical Intelligence, Lovable, Chai Discovery, Reducto, Substack) and larger enterprises (Quora), with documented outcomes including 65% latency reduction (Decagon), 3x latency decrease (Reducto), 10-15 ms latency for robot control (Physical Intelligence), and 4 months faster time-to-launch (Suno). The platform is SOC2 Type 2 and HIPAA compliant, has reported 99.937% uptime over a 30-day window, and supports sub-10ms overhead latency for online inference from globally distributed compute. Modal's customer base is horizontal, and marketing and developer relations are built around documentation, an open-source GitHub examples repository, a Slack community, technical blog posts, a GPU Glossary, and YouTube tutorials.

Name

Modal Labs

Legal name

Modal Labs

Website

https://modal.com

Company type

Private

Founded year

2021

Operating status

Operating

Headcount range

51–100 employees

Short description

Modal Labs provides a serverless GPU cloud platform purpose-built for AI workloads, offering inference, training, batch processing, and sandboxed code execution via a Python SDK. The platform serves AI/ML developers, AI-native startups (Suno, Runway, Decagon, Lovable), and enterprises (Quora) with usage-based pricing and self-service onboarding.

Ownership category

akta.pro rank

Product category

Cloud GPU Infrastructure

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518), Computer Facilities Management Services (541513), Computer Systems Design and Related Services (54151)

SIC

Services-Prepackaged Software (7372), Services-Computer Processing & Data Preparation (7374)

akta.pro industry

PRIMARYHDAAAAAK

AI Compute Cloud & GPU-as-a-Service

Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)

SECONDARYHDABAAAC

Serverless & Functions-as-a-Service (FaaS)

Hardware, Technology & Telecom/Cloud & Data Center Infrastructure/Public Cloud Infrastructure (IaaS/PaaS)

SECONDARYHDAAAAAG

AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers)

Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/AI Infrastructure & Compute (GPUs/ASICs, accelerators)

SECONDARYHDACABAG

GPU-Accelerated & AI Training/Inference Servers

Hardware, Technology & Telecom/Computing Hardware & Peripherals/Servers & Enterprise Compute Systems

Keywords

• Serverless GPU cloud
• AI infrastructure platform
• GPU compute orchestration
• ML training and inference
• AI agent runtime

Headquarters

HQ city

New York

HQ country

United States

HQ region

North America

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Infrastructure, Personnel, Marketing or Sales, Operations

Revenue model

1. GPU compute consumption: Pay-per-second usage for GPU compute time. Users scale from 0 to 1000+ GPUs and pay only for what they use. No reserved capacity required.
2. Free tier / trial credit: $30 free compute credit for new users to try the platform. Enables product-led onboarding without upfront commitment.

Pricing tiers

Go-to-market motion1 record

Distribution channels1 record

Marketing channels6 records

Core offering

Modal Labs operates an AI-native serverless GPU cloud platform that lets developers run inference, training, batch processing, and isolated code execution on NVIDIA GPUs (H100s, A100s, B200s, H200s, L40S, A10Gs) without managing infrastructure. Workloads are defined in Python via the Modal SDK, deploy in seconds, and elasticly scale from 0 to 1,000+ GPUs across multiple cloud providers with sub-second cold starts.

Product overview

Modal Labs offers an AI-native cloud platform for serverless GPU workloads. The Core Platform provides the foundational runtime with sub-second cold starts, instant autoscaling (0 to 1000+ GPUs), and GPU infrastructure (H100s, A100s, A10Gs, B200s, H200s, L40S). Built on top are three core products: Inference (LLM and multi-modal inference serving), Training (fine-tuning, multi-node training, RL, hyperparameter sweeps), and Sandboxes (isolated code execution for agents and untrusted code). Add-on modules include Batch (parallel batch inference), Notebooks (Jupyter on GPUs), and the Modal SDK (Python developer interface). The entire portfolio is programmable from Python with no YAML, no infrastructure configuration files, and scales to zero between requests. Pricing offers $30/month free compute, with GPU workloads billed per second.

Differentiator

Problem solved

Functional benefit

Brands

• Modal SDK: Python SDK for cloud compute infrastructure
• Modal Cloud
• DoppelBot
• GPU Glossary

Products and services

• Inference Serverless inference product for deploying and scaling LLM and multi-modal inference workloads (text, image, video, audio, embeddings). Supports any model or inference engine on H100s, A100s, A10Gs and more, with scale-to-zero between requests and burst scaling. Includes OpenAI-compatible API endpoints, token streaming, WebRTC, and WebSocket support.
• Training End-to-end training product for fine-tuning and training ML models, from single-GPU SFT, LoRA, and full fine-tunes to multi-node clusters up to 128 B200s with 3200 Gbps Infiniband networking. Supports reinforcement learning with thousands of concurrent trajectories and parallel hyperparameter sweeps.
• Sandboxes Isolated, ephemeral, GPU-accelerated execution environments for running untrusted code, coding agents, background agents, and RL rollouts. Spins up hundreds of thousands of concurrent rollout environments in seconds with custom images and dependencies.
• Batch

Quantifiable outcome

• 65% latency reduction for Decagon
• +5 more outcomes

Named customers9 records

Segments4 records

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration10 records

AI capability11 records

Feature9 records

Partnerships

Four partnerships are on record, tiered core.

• Hugging FacecoreTechnology or IntegrationModal integrates with Hugging Face Hub for model hosting and downloads. Modal Volumes are used for caching Hugging Face model weights. Users can fine-tune models from Hugging Face and deploy inference endpoints.
• vLLMcoreTechnology or IntegrationModal uses vLLM as the primary inference engine for LLM serving, with native support for LoRA adapter swapping and OpenAI-compatible APIs.
• AWS S3coreTechnology or IntegrationModal CloudBucketMount feature enables mounting S3 buckets directly into Modal apps for reading and writing cloud storage as local directories.
• NVIDIAcoreTechnology or IntegrationModal provides access to NVIDIA GPUs (H100s, A100s, A10Gs, B200s, H200s, L40S) across cloud providers. GPU scheduling and fleet management built on NVIDIA infrastructure.

Scale indicators5 records

Recent moves6 records

Expansion highlights7 records

Direct peers

• CoreWeave: CoreWeave is a GPU-focused cloud provider offering large-scale H100/B200 compute for AI training and inference. It is the closest direct peer to Modal — same customer base (AI labs, startups, enterprises), same product (GPU cloud), competing on performance, scale, and developer tooling.
• Lambda Labs: Lambda operates GPU cloud instances and clusters for AI training and inference, with both on-demand and reserved offerings. Competes with Modal for AI developers seeking high-performance GPU compute, with a similar focus on Python/ML workflows.
• Together AI: Together AI provides serverless and dedicated GPU inference for open-source LLMs, with a developer-friendly API and per-token pricing. Closely parallels Modal's Inference product in targeting AI startups with pay-as-you-go GPU inference.
• RunPod: RunPod offers on-demand GPU pods and serverless endpoints for AI inference and training at competitive prices. Targets the same developer-led AI workload segment as Modal and competes on price and per-second billing.
• Anyscale: Anyscale (creators of Ray) provides managed compute for distributed AI workloads, including training, fine-tuning, and inference at scale. Competes with Modal's Training product for ML teams running multi-node training and RL rollouts.
• Replicate: Replicate offers a serverless cloud for running open-source ML models via API, with a developer-focused, PLG motion similar to Modal. Targets AI developers wanting simple per-prediction inference without managing infrastructure.

Emerging players

• Fireworks AI: Fireworks AI provides optimized open-source LLM inference and fine-tuning APIs with low-latency serving. Comparable to Modal's Inference and LoRA-swap capabilities, with stronger emphasis on model API rather than raw GPU infrastructure.

Broad incumbents

• AWS (SageMaker / Bedrock): AWS offers managed AI/ML infrastructure (SageMaker) and foundation-model services (Bedrock) with broader enterprise reach. Competes with Modal across the same workloads but lacks Modal's AI-native developer ergonomics and sub-second cold-start focus.
• Google Cloud Vertex AI: Google Cloud's Vertex AI provides managed training, tuning, and inference on TPUs/GPUs with deep integration into the Google ecosystem. Competes broadly with Modal for AI workloads, particularly for customers already on GCP.
• Microsoft Azure AI: Azure AI offers managed compute, model serving, and MLOps tooling integrated with the broader Azure enterprise stack. Competes with Modal for enterprise AI workloads, particularly via OpenAI partnership and Azure ML.

Market position

Strengths5 records

Weaknesses4 records

Competitive moat5 records

Key risks5 records

Key highlights7 records

Customer concentration

• GitHubgithub.com/modal-labs
• Xx.com/modal_labs
• LinkedInlinkedin.com/company/modal-labs

Compliance2 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles1 record

Funding overview

Funding rounds5 records

Investors19 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A2 records

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Modal Labs