Latest stories
Modal's latest updates feature faster memory snapshotting, expanded async job queues, and OpenID Connect for improved authentication and integration.
Modal Sandboxes are now generally available, offering a secure environment for executing untrusted code with dynamic configurations and robust functionality.
The NVIDIA L40S GPUs are now available on Modal, offering enhanced memory, speed, and several new features for developers and researchers.
Modal's introduction of memory snapshots significantly decreases cold start latency in GPU container runtimes, offering boosts in performance over traditional a...
This guide explores GPU utilization metrics, focusing on allocation, kernel, and FLOP/s utilization, providing insights on achieving optimal performance in mach...
Modal has slashed hardware prices by 65%, launched a new SDK on conda-forge, and provides substantial updates for developers tackling GPU challenges.
Modal has refreshed its logo and rolled out its first out-of-home campaign in San Francisco, promoting its AI infrastructure platform with targeted messaging.
Modal introduces capabilities for batch processing with 1M inputs, improved client updates, and a fast LLM inference example using TensorRT-LLM.
sync. leverages Modal to efficiently lipsync over 100 hours of video daily, enhancing their AI-driven editing capabilities while streamlining operations.
Modal's resource solver utilizes linear programming to optimize GPU allocation, addressing volatile cloud compute pricing by providing robust, scalable solution...
Modal's Dicts now offer unlimited storage, enhanced caching, and distributed locking, significantly improving performance for developers managing state in conta...
Modal Batch introduces an enhanced job processing system allowing seamless handling of up to 1 million tasks, significantly improving efficiency for AI workload...
Twirl's integration into Modal aims to bolster data orchestration and serverless infrastructure in the growing European AI market.
Modal introduces B200 and H200 GPUs, offering significant performance gains for large models at accessible pricing without complex sales processes.
This article details how to enhance the FLUX.1-dev model's performance, achieving threefold speed improvements and reduced latency in generative image processin...
Learn how innovative techniques like evals and compute scaling improved the functionality and aesthetics of generated QR codes, achieving a 95% scan rate.
Modal's acquisition of Jamsocket aims to integrate advanced capabilities for developing AI-centric collaborative applications on a unified platform.
Modal introduces multi-node training clusters, new NVIDIA B200 and H200 servers, and updates its client, enhancing efficiency for developers.
Recent advancements in open automated speech recognition models offer significant cost and efficiency improvements for transcription tasks.
Modal's introduction of GPU memory snapshots accelerates cold start times for GPU workloads, achieving startup speeds up to 10x faster than traditional methods.