Modal Blog
How Reducto improved enterprise-scale document processing latency by 3x
Learn how Reducto used GPU memory snapshotting and flexible autoscaling to build fast multi-model pipelines.
Host overhead is killing your inference efficiency
Never block the GPU.
How Decagon shipped real-time voice AI on Modal
How Decagon and Modal made real-time voice AI possible, combining fine-tuned small models with a re-engineered inference runtime for sub-second latency.
Latest in Engineering
Host overhead is killing your inference efficiency
Never block the GPU.
One-Second Voice-to-Voice Latency with Modal, Pipecat, and Open Models
How we built a real-time voice bot on Modal's distributed serverless platform.
We reverse-engineered Flash Attention 4
Asynchrony, fast approximate exponents, and 10x more efficient softmax.
Latest in News
Product Updates: Updates to Volumes, JS and Go SDKs, and more
Welcome to another round of Modal Product Updates! Here's what's new this month.
Modal + Datalab: Deploy high-throughput document intelligence in <5 minutes
We've collaborated with Datalab, the creators of Marker and Surya, to make it faster than ever to deploy document intelligence workflows.
Announcing our $87M Series B
We’re excited to announce that we have raised more than $80M in a Series B round, led by Lux Capital. Our post-money valuation is $1.1B.
Latest in Customer Stories
How Reducto improved enterprise-scale document processing latency by 3x
Learn how Reducto used GPU memory snapshotting and flexible autoscaling to build fast multi-model pipelines.
How Decagon shipped real-time voice AI on Modal
How Decagon and Modal made real-time voice AI possible, combining fine-tuned small models with a re-engineered inference runtime for sub-second latency.
How Zencastr transcribed hundreds of years worth of audio in just a few days
Zencastr scaled up to 1,500 concurrent GPUs on Modal to process hundreds of years of podcast audio in just a few days. Today they run transcription, speaker detection, and audio enrichment for millions of podcast episodes on Modal, giving them cost efficiency, fast iteration, and zero DevOps overhead.
'/%3e%3cpath%20d='M109.623%2064L73.2925%201.07001C72.0925%201.76001%2071.0825%202.76%2070.3625%204L1.0725%20124C-0.3575%20126.48%20-0.3575%20129.52%201.0725%20132L33.4026%20188C34.1126%20189.24%2035.1325%20190.24%2036.3325%20190.93L109.613%2064H109.623Z'%20fill='url(%23paint1_linear_342_139)'/%3e%3cpath%20d='M183.513%2064H109.613L36.3325%20190.93C37.5325%20191.62%2038.9025%20192%2040.3325%20192H104.993C107.853%20192%20110.492%20190.47%20111.922%20188L183.513%2064Z'%20fill='%2309AF58'/%3e%3cpath%20d='M365.963%20132C366.673%20130.76%20367.033%20129.38%20367.033%20128H294.372L258.042%20190.93C259.242%20191.62%20260.612%20192%20262.042%20192H326.703C329.563%20192%20332.202%20190.47%20333.632%20188L365.963%20132Z'%20fill='%2309AF58'/%3e%3cpath%20d='M225.083%200C223.653%200%20222.283%200.380007%20221.083%201.07001L294.362%20128H367.023C367.023%20126.62%20366.663%20125.24%20365.953%20124L296.672%204C295.242%201.53%20292.603%200%20289.743%200H225.073H225.083Z'%20fill='url(%23paint2_linear_342_139)'/%3e%3cpath%20d='M258.033%20190.93L294.362%20128L221.083%201.07001C219.883%201.76001%20218.873%202.76%20218.153%204L183.513%2064L255.103%20188C255.813%20189.24%20256.833%20190.24%20258.033%20190.93Z'%20fill='url(%23paint3_linear_342_139)'/%3e%3cdefs%3e%3clinearGradient%20id='paint0_linear_342_139'%20x1='155.803'%20y1='80'%20x2='101.003'%20y2='-14.93'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20stop-color='%23BFF9B4'/%3e%3cstop%20offset='1'%20stop-color='%2380EE64'/%3e%3c/linearGradient%3e%3clinearGradient%20id='paint1_linear_342_139'%20x1='8.62251'%20y1='174.93'%20x2='100.072'%20y2='16.54'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20stop-color='%2380EE64'/%3e%3cstop%20offset='0.18'%20stop-color='%237BEB63'/%3e%3cstop%20offset='0.36'%20stop-color='%236FE562'/%3e%3cstop%20offset='0.55'%20stop-color='%235ADA60'/%3e%3cstop%20offset='0.74'%20stop-color='%233DCA5D'/%3e%3cstop%20offset='0.93'%20stop-color='%2318B759'/%3e%3cstop%20offset='1'%20stop-color='%2309AF58'/%3e%3c/linearGradient%3e%3clinearGradient%20id='paint2_linear_342_139'%20x1='340.243'%20y1='143.46'%20x2='248.793'%20y2='-14.93'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20stop-color='%23BFF9B4'/%3e%3cstop%20offset='1'%20stop-color='%2380EE64'/%3e%3c/linearGradient%3e%3clinearGradient%20id='paint3_linear_342_139'%20x1='284.822'%20y1='175.47'%20x2='193.372'%20y2='17.0701'%20gradientUnits='userSpaceOnUse'%3e%3cstop%20stop-color='%2380EE64'/%3e%3cstop%20offset='0.18'%20stop-color='%237BEB63'/%3e%3cstop%20offset='0.36'%20stop-color='%236FE562'/%3e%3cstop%20offset='0.55'%20stop-color='%235ADA60'/%3e%3cstop%20offset='0.74'%20stop-color='%233DCA5D'/%3e%3cstop%20offset='0.93'%20stop-color='%2318B759'/%3e%3cstop%20offset='1'%20stop-color='%2309AF58'/%3e%3c/linearGradient%3e%3c/defs%3e%3c/svg%3e)