Bernt Børnich, founder and CEO of 1X, is joining us at Runtime. His deep expertise in machine learning and autonomous systems, combined with a strong user-centered design philosophy, has shaped everything, from early prototypes to the NEO Home Robot. He'll talk through the making of NEO and his bet on a more abundant, human-centered future.
Modal
Software Development
New York City, New York 32,582 followers
AI needs a new infrastructure layer. We're building it.
About us
Customers rely on Modal for instant GPU access, sub-second container starts, and native storage, so it's simple to serve low-latency inference, fine-tune models, and access production-ready sandboxes at scale. Every era of computing came with new workloads that previous infrastructure couldn't serve: mainframes, databases, the cloud. Each time, the company that rebuilt the layer underneath defined the decade. AI is no different, except it touches everything instead of one slice. The window to build is open right now.
- Website
-
https://modal.com
External link for Modal
- Industry
- Software Development
- Company size
- 51-200 employees
- Headquarters
- New York City, New York
- Type
- Privately Held
- Specialties
- Serverless GPUs, LLM Inference, LLM Fine-Tuning, Generative Model Inference, Generative Model Training, Computational Biology, Audio Generation, Image Generation, Video Generation, Web Scraping, Batch Jobs, Batch Embeddings, Scaling Out, AI Agents, Reinforcement Learning, Sandboxes, and Background Agents
Products
![Click here to view Modal]()
Modal
Platform as a Service (PaaS) Software
Modal is a serverless compute platform that makes it easy for developers to run compute-intensive workloads like ML inference, fine-tuning, and batch jobs. Our proprietary Rust-based container stack is best-in-class, allowing you to run any function in the cloud in less than a second, even on the most in-demand GPU types. We autoscale to thousands of GPUs or CPUs for your functions based on request volume so you can always meet customer demand while never paying for idle resources. Modal's Python SDK allows you to define custom images and hardware requirements in code. No more spending time on config files or cloud consoles. Let your team ship innovative AI products—we'll handle the compute.
Locations
-
Primary
Get directions
New York City, New York 10038, US
-
Get directions
Stockholm , SE
-
Get directions
San Francisco, California 94103, US
Employees at Modal
Updates
-
Join Diogo Almeida, Co-founder and CEO of TypeSafe AI at Runtime on October 1st. Apply to attend: modal.com/runtime
-
We've partnered with Anthropic to contribute up to $250K in Modal credits for researchers pushing protein design forward in the open 🧬
Biologists use specialized open-source models for tasks like modeling the structure of molecular systems, designing drug-like molecules, and predicting the effects of genetic mutations. But these models are often expensive to run, potentially limiting their impact. In our latest Science Blog, we share how Claude was able to optimize inference for more than 30 open-source models, making them 4x faster on average, partly by writing custom software for GPUs. We’re open sourcing all of the optimization code. Read more here: https://lnkd.in/gjg7ACnz To show what these optimizations make possible, we’re partnering with Adaptyv Bio on a protein design competition. Together, we’ll be experimentally validating over 5,000 designs. We're providing up to $1 million in Claude credits plus funding alongside Adaptyv for experimental validation. Modal is contributing up to $250,000 in compute and Twist Bioscience is providing DNA. Learn more on Adaptyv’s Proteinbase: https://lnkd.in/ddwgnVE4 And sign up for the competition here: https://lnkd.in/g5v6-cQQ You can find all of the code on GitHub: https://lnkd.in/gPgDEDPV And the full results in our technical report: https://lnkd.in/gwdtN2-z
-
Modal reposted this
We're opening up more spots! • October 3-4th • 150 builders • Win $5,000+ in cash & compute • Get 1-1 mentorship from Anthropic, Google DeepMind, Hugging Face and our partners: Originator • Modal • IQ Capital • Cambridge Compute Co. • Voyager Ventures • GXL • Amass • AIDDA Institute Applications still open here: https://luma.com/3iipivod
-
-
Ship, scale, and sail with Modal, MotherDuck, and Braintrust!
Our last boat party filled immediately.. SO we're doing it again. This time with Braintrust and Modal during TECH WEEK by a16z. If you're an SF based engineer, data, or AI leader, come Ship, Scale, and Sail with us! https://lnkd.in/emd-x6EK
-
-
Paris, we're on our way 🇫🇷 Next week the Modal team flies in from New York, San Francisco, London and Stockholm for AI Engineer Paris (September 23-24). If you're going, we'd love to meet you. Find us at the booth both days, or catch Charles Frye's talk on optimizing low-latency inference on Thursday at 10:00 CET. And if you're in town the night before, join us for warm-up drinks with Black Forest Labs and H Company. Sign up via link below.
-
-
Modal reposted this
Can AI agents invent a language of their own, one humans might not understand? We built GlossoGen an open-source platform where groups of AI models have to talk to each other to solve problems: treating a patient in a simulated emergency, or playing spot-the-difference across two scenes. Every model starts out in plain English. But put pressure on the conversation, a need for speed, say, or a channel that randomly drops characters, and something remarkable happens. New languages emerge. First new words, then new grammar, until no human can follow what the agents are saying. And they were never asked to hide anything. The opacity emerged on its own. One finding stands out. Only the most capable recent models could invent a new language. But even the models that couldn't invent one quickly learned a language other agents had created. New languages are hard to originate. Once they exist, they spread. That is the point of this work: to understand how and why AI agents create new languages, how those languages evolve, and what that means for our ability to oversee systems built from many interacting agents, while the conversation is still one humans can follow. This research is part of the Schmidt Sciences AI Agents Evolving Communication and Coordination pilot program, led by Elias Stengel-Eskin (UT Austin) and Simon Kirby (University of Edinburgh), supported by AE Studio. GlossoGen is open source and supports both closed model APIs and open-source models running on Modal.
-
-
Modal reposted this
AI Engineer Paris is just one week away! This second edition of AI Engineer is promising to be one for the books. Main Stage Keynotes from leading engineers at Mistral, Hugging Face, Black Forest Labs, Google DeepMind, Arm, and many more of Europe’s most exciting AI companies. A huge thank you to all our incredible sponsors for making this special event possible: NVIDIA Factory Sentry Cockroach Labs Sonar Sierra Notion Backblaze Cloudflare Qdrant Modal Langfuse Daytona Neo4j Stripe CodeRabbit Descope ZenML Makora Latitude AI Cominty AI Fimo.ai Regular Bird tickets sold out. Late Bird tickets are almost sold out. Get yours before they’re gone. 🎟️
-
Runtime speaker lineup is live! We're bringing together experts covering AI infrastructure, applications of AI in science and robotics, the future of software engineering, and more.
The conference for engineers running AI in production
-
Frontier models are simply too expensive and slow for the majority of use cases, so we see models like SWE-2 becoming the daily driver for most. Training trillion-parameter coding agents at scale isn't easy though: typically, each step launches thousands of rollouts, each with its own isolated environment. Cognition uses Modal's sandbox infrastructure for the rollouts behind SWE-2. Congrats on the launch! More on how we scale Sandboxes in the comments.
Introducing SWE-2, our closest model yet to the frontier. On leading coding evals, SWE-2 scores on par with recent frontier models at up to 70% lower cost. On FrontierCode 1.1, it matches Fable 5.1 at 64% lower cost. It's our first RL run at multi-trillion parameter scale, trained with a refined recipe that pushes the Pareto curve of capability and cost. A few things we're proud of: - Effort levels. SWE-2 is our first model to support them. In a single RL run, medium effort gets cheaper and smarter, while max effort learns to use more tokens and turns for the highest scores. - Efficiency. On FrontierCode 1.1, SWE-2 medium scores higher than SWE-1.7 with 58% fewer turns and 81% lower cost. It explores more deliberately before making changes. - Real work. Our team uses SWE-2 daily in Devin for feature development, debugging, and complex visualizations. SWE-2 is available today in Devin across Desktop and CLI, and free for all Pro, Max & Teams subscribers for the next month.
-