What Modal Builds
Modal provides serverless compute infrastructure designed specifically for AI engineering workloads—combining fast-booting GPU containers for inference and RL/fine-tuning with isolated Sandboxes for executing coding agents and untrusted code. On modal.com and modal.com/docs, engineers define container images, GPU attachments, volumes, and web endpoints directly in Python, JavaScript, or Go without maintaining Kubernetes clusters.
Modal also serves hosted open-weight Model APIs (such as Kimi K3.1 Pro, DeepSeek V4.1 Flash, Qwen3.8-Max, and GLM 5.3) billed per million tokens alongside its core container runtime, and publishes engineering benchmarks demonstrating scaling to one million concurrent sandboxes.
Quick-Reference Fact Sheet
| Field | Verified Record (As of November 20, 2026) |
|---|---|
| Company | Modal |
| Official Website | https://modal.com/ |
| Primary Category / Role | Serverless GPU cloud, AI inference/training runtime & large-scale coding-agent Sandboxes |
| Census Cluster | Agent Engineering, Observability & Infrastructure |
| Operating Status | Operating — general availability (per-second serverless cloud) |
| Independence Gate | Independent private company |
| Published Pricing Summary | Starter ($0 + per-second compute, $30/mo free credits); pay-per-second GPU (T4 $0.000164/s to B300 $0.001972/s) & CPU ($0.0000131/core/s) |
| Key Protocols & Surfaces | Python, JavaScript, and Go SDKs, Modal Sandboxes, OpenAI-compatible Model APIs, serverless GPUs, Volumes |
| CSV Population Rank | #41 (Chunk 1 of AI Dev-Tools Census) |
| Verified Record Date | November 20, 2026 |
Who It Is For and How It Works
Modal targets AI infrastructure engineers, ML researchers, and coding-agent teams that need sub-second elastic scaling across CPUs and GPUs without paying for idle cluster capacity. Because compute bills strictly per second of execution, bursty workloads—such as parallel RL rollouts, batch evaluation harnesses, or on-demand coding sandboxes—scale to zero automatically.
Developers start on the Starter plan ($0 base fee, $30/month free credits, 3 workspace seats, 100 containers, and 10 concurrent GPUs) and scale into team and enterprise tiers with dedicated capacity and compliance controls.
- Four core pillars: Serverless AI Inference, isolated Agent Sandboxes, on-demand GPU Training/RL, and hosted Model APIs.
- True per-second metering: Zero charges for idle capacity; granular billing across T4, L4, A10, L40S, A100, H100, H200, B200, and B300 GPUs.
- Multi-language SDKs: Python, JavaScript, and Go SDKs with programmatic container definitions and persistent Volumes.
- Included Starter allowance: $30/month free compute credits, 3 workspace seats, and 1 TiB/month free Volume allowance.
Published Pricing and Commercial Availability
On modal.com/pricing (verified November 20, 2026), Modal lists Starter at $0/month plus compute with $30/month in free credits (3 seats, 100 containers, 10 GPU concurrency). Published per-second rates include CPU at $0.0000131/core/sec, Memory at $0.00000222/GiB/sec, Volumes at $0.09/GiB/month (after 1 TiB free), Network Egress at $0.04/GiB, and GPUs ranging from Nvidia T4 ($0.000164/sec) and H100 SXM5 ($0.001097/sec) up to Nvidia B300 ($0.001972/sec), plus per-token Model APIs.
Corporate Independence and Funding Context
Modal Labs operates as an independent private company. Primary product and company pages at modal.com do not publish private valuation tables.
Related AI Dev-Tools Census Profiles
- AI Dev-Tools Census Index — Methodology, 85-Company Gate & Chunk 1 Directory
- E2B — AI Dev-Tools Census Profile (Agent Engineering, Observability & Infrastructure)
- Daytona — AI Dev-Tools Census Profile (Agent Engineering, Observability & Infrastructure)
- LangChain — AI Dev-Tools Census Profile (Agent Engineering, Observability & Infrastructure)
Verified Primary Sources
- Modal Official Website (https://modal.com/)
- Modal Pricing (https://modal.com/pricing)
- Modal Documentation (https://modal.com/docs)
- Modal Company (https://modal.com/company)
Get The Signal
What we actually ran, what it measured, and what it means for a restoration shop — every week, including the flops.

Leave a Reply