{"slug": "modal", "name": "Modal", "category": "infra", "category_name": "Infrastructure and DevOps", "description": "Serverless cloud platform for running Python code, GPU jobs and sandboxes without managing infrastructure.", "recommenders": 6, "commenters": 11, "warners": 1, "score": 5.374, "rising": -0.64, "first": "2025-11-17", "latest": "2026-08-31", "why": "Recommended for serverless GPU workloads, with a well-designed SDK and faster setup than provisioning instances by hand, and listed among sandbox providers for running AI agents.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 6, "neutral": 2, "compare": 2, "warn": 1}, "spark": [1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0], "listed": true, "loved": ["GPU serverless", "Agent sandboxes"], "caveat": "Docker in sandboxes", "short_quote": "I recommend the service but like I said I haven't tried any others.", "icon": true, "domain": null, "github_repo": null, "official_url": "https://modal.com", "pricing": "freemium", "aliases": ["Modal"], "verdict": [{"text": "Recommended for serverless GPU workloads, with a well-designed SDK and faster setup than provisioning instances by hand,", "cites": [{"id": 753368656, "item_id": 47085541, "url": "https://news.ycombinator.com/item?id=47085541", "by": "isoprophlex", "date": "2026-02-20", "stance": "recommend", "quote": "I love modal. I think they got FaaS for GPU exactly right", "excerpt": "> There are no great FaaS options for running GPU workloads I love modal. I think they got FaaS for GPU exactly right, both in terms of their SDK and the abstractions/infra they provide.", "context": "", "use_case": "GPU FaaS workloads", "switched_from": "", "story": "Almost Every infrastructure decision I endorse or regret after 4 years"}, {"id": 792237824, "item_id": 49514864, "url": "https://news.ycombinator.com/item?id=49514864", "by": "ronfriedhaber", "date": "2026-08-31", "stance": "recommend", "quote": "Modal significantly improves this. Highly recommend.", "excerpt": "> Just provisioning a new instance and copying the relevant checkpoints and things can take 25 minutes. Modal significantly improves this. Highly recommend.", "context": "", "use_case": "provisioning ML training instances", "switched_from": "", "story": "Apple caught off guard by AI demand for Mac Mini and Mac Studio"}, {"id": 747584592, "item_id": 46724037, "url": "https://news.ycombinator.com/item?id=46724037", "by": "dataviz1000", "date": "2026-01-22", "stance": "recommend", "quote": "I recommend the service but like I said I haven't tried any others.", "excerpt": "Hey, I use Modal! I don't know how it compares to any other service because I used it once and it worked perfect at price point I found reasonable. Also, working with Claude code and the Modal cli API is perfect for managing all the jobs and moving assets around. I tried writing scripts to manager assets across my local device and Google's colab but it was still not ideal. So two months ago I asked chatgpt to search ", "context": "light user, $20-40/month", "use_case": "ML training jobs on GPUs", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}]}, {"text": " and listed among sandbox providers for running AI agents.", "cites": [{"id": 779023829, "item_id": 48688989, "url": "https://news.ycombinator.com/item?id=48688989", "by": "dbmikus", "date": "2026-06-26", "stance": "recommend", "quote": "Daytona, E2B, OpenComputer, Freestyle, Blaxel, Vercel, Modal, Cloudflare, Tensorlake, Superserve, etc. etc.", "excerpt": "Why do you want to self-host vs. using one of the many providers out there? Daytona, E2B, OpenComputer, Freestyle, Blaxel, Vercel, Modal, Cloudflare, Tensorlake, Superserve, etc. etc. Some of them work by pre-purchasing credits, so you can control the blast radius of spend. Also, if you want a more embedded sandbox runtime as a library instead of a daemon + REST API, you can check out libkrun (and friendly layers on ", "context": "", "use_case": "hosted sandboxes", "switched_from": "", "story": "Run isolated sandboxes with full lifecycle control: AWS introduces MicroVMs"}, {"id": 736237299, "item_id": 46014831, "url": "https://news.ycombinator.com/item?id=46014831", "by": "mritchie712", "date": "2025-11-22", "stance": "recommend", "quote": "We use e2b.dev, but Modal is great too.", "excerpt": "Some things we've[0] learned on agent design: 1. If your agent needs to write a lot of code, it's really hard to beat Claude Code (cc) / Agent SDK. We've tried many approaches and frameworks over the past 2 years (e.g. PydanticAI), but using cc is the first that has felt magic. 2. Vendor lock-in is a risk, but the bigger risk is having an agent that is less capable then what a user gets out of chatgpt because you're ", "context": "", "use_case": "agent sandbox computer", "switched_from": "", "story": "Agent design is still hard"}]}, {"text": " Complaints are isolated: one commenter says its gVisor sandboxes make running Docker inside harder,", "cites": [{"id": 786594898, "item_id": 49162181, "url": "https://news.ycombinator.com/item?id=49162181", "by": "dregitsky", "date": "2026-08-03", "stance": "compare", "quote": "Modal, but they use gVisor and not Firecracker, which means it's harder for users to run docker inside the box", "excerpt": "For boxes.dev we use E2B - they had everything we needed at the time. I'd love a standardized API / marketplace, but more just to try out new providers with differentiated features or pricing - we probably wouldn't be switching often / using multiple (unless customers asked us for more options). There are so many sandbox providers popping up! Our use case is pretty specific though - one persistent machine that you se", "context": "builds boxes.dev", "use_case": "sandbox VMs for agents", "switched_from": "", "story": "Devtools must be open source"}]}, {"text": " and another doubts that GPU-as-a-service middlemen like Modal are sustainable.", "cites": [{"id": 747583504, "item_id": 46723969, "url": "https://news.ycombinator.com/item?id=46723969", "by": "bflesch", "date": "2026-01-22", "stance": "warn", "quote": "business model of GPU-as-a-service middleman companies like modal deeply unsustainable", "excerpt": "In his newsletter Ed Zitron hammered down the point that GPUs depreciate quickly, but these kind of reliability issues are shocking to read. The GPUs are so common to fail that they hang out in a 24/7 slack channel with customers like Meta (who apparently can't set up a cluster themselves..). Ed Zitron also called out the business model of GPU-as-a-service middleman companies like modal deeply unsustainable, and I al", "context": "", "use_case": "GPU cloud", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}]}], "praise": [{"label": "GPU serverless", "summary": "Praised for getting function-as-a-service for GPU workloads right and for cutting slow instance setup.", "quotes": [{"id": 753368656, "item_id": 47085541, "url": "https://news.ycombinator.com/item?id=47085541", "by": "isoprophlex", "date": "2026-02-20", "stance": "recommend", "quote": "I love modal. I think they got FaaS for GPU exactly right", "excerpt": "> There are no great FaaS options for running GPU workloads I love modal. I think they got FaaS for GPU exactly right, both in terms of their SDK and the abstractions/infra they provide.", "context": "", "use_case": "GPU FaaS workloads", "switched_from": "", "story": "Almost Every infrastructure decision I endorse or regret after 4 years"}, {"id": 792237824, "item_id": 49514864, "url": "https://news.ycombinator.com/item?id=49514864", "by": "ronfriedhaber", "date": "2026-08-31", "stance": "recommend", "quote": "Modal significantly improves this. Highly recommend.", "excerpt": "> Just provisioning a new instance and copying the relevant checkpoints and things can take 25 minutes. Modal significantly improves this. Highly recommend.", "context": "", "use_case": "provisioning ML training instances", "switched_from": "", "story": "Apple caught off guard by AI demand for Mac Mini and Mac Studio"}, {"id": 788183125, "item_id": 49261445, "url": "https://news.ycombinator.com/item?id=49261445", "by": "yummyelephant8", "date": "2026-08-11", "stance": "neutral", "quote": "hooked it up to vLLM on Modal", "excerpt": "Like you, I use coding agents like Claude Code, Codex, and Cursor every day. But the other day, I was curious about how Cursor trained their original tab-completion model. So I started wondering: could you post-train an open-source LLM to autocomplete ComfyUI workflows given your current workflow, similar to how Composer autocompletes code? More specifically: Could I post-train an LLM to predict up to four nodes in t", "context": "", "use_case": "inference hosting", "switched_from": "", "story": null}]}, {"label": "Agent sandboxes", "summary": "Named alongside E2B and others as a good hosted sandbox option for AI agents.", "quotes": [{"id": 736237299, "item_id": 46014831, "url": "https://news.ycombinator.com/item?id=46014831", "by": "mritchie712", "date": "2025-11-22", "stance": "recommend", "quote": "We use e2b.dev, but Modal is great too.", "excerpt": "Some things we've[0] learned on agent design: 1. If your agent needs to write a lot of code, it's really hard to beat Claude Code (cc) / Agent SDK. We've tried many approaches and frameworks over the past 2 years (e.g. PydanticAI), but using cc is the first that has felt magic. 2. Vendor lock-in is a risk, but the bigger risk is having an agent that is less capable then what a user gets out of chatgpt because you're ", "context": "", "use_case": "agent sandbox computer", "switched_from": "", "story": "Agent design is still hard"}, {"id": 779023829, "item_id": 48688989, "url": "https://news.ycombinator.com/item?id=48688989", "by": "dbmikus", "date": "2026-06-26", "stance": "recommend", "quote": "Daytona, E2B, OpenComputer, Freestyle, Blaxel, Vercel, Modal, Cloudflare, Tensorlake, Superserve, etc. etc.", "excerpt": "Why do you want to self-host vs. using one of the many providers out there? Daytona, E2B, OpenComputer, Freestyle, Blaxel, Vercel, Modal, Cloudflare, Tensorlake, Superserve, etc. etc. Some of them work by pre-purchasing credits, so you can control the blast radius of spend. Also, if you want a more embedded sandbox runtime as a library instead of a daemon + REST API, you can check out libkrun (and friendly layers on ", "context": "", "use_case": "hosted sandboxes", "switched_from": "", "story": "Run isolated sandboxes with full lifecycle control: AWS introduces MicroVMs"}]}, {"label": "SDK and CLI", "summary": "Commenters like its SDK abstractions and say the CLI works well for managing jobs and assets.", "quotes": [{"id": 753368656, "item_id": 47085541, "url": "https://news.ycombinator.com/item?id=47085541", "by": "isoprophlex", "date": "2026-02-20", "stance": "recommend", "quote": "I love modal. I think they got FaaS for GPU exactly right", "excerpt": "> There are no great FaaS options for running GPU workloads I love modal. I think they got FaaS for GPU exactly right, both in terms of their SDK and the abstractions/infra they provide.", "context": "", "use_case": "GPU FaaS workloads", "switched_from": "", "story": "Almost Every infrastructure decision I endorse or regret after 4 years"}, {"id": 747584592, "item_id": 46724037, "url": "https://news.ycombinator.com/item?id=46724037", "by": "dataviz1000", "date": "2026-01-22", "stance": "recommend", "quote": "I recommend the service but like I said I haven't tried any others.", "excerpt": "Hey, I use Modal! I don't know how it compares to any other service because I used it once and it worked perfect at price point I found reasonable. Also, working with Claude code and the Modal cli API is perfect for managing all the jobs and moving assets around. I tried writing scripts to manager assets across my local device and Google's colab but it was still not ideal. So two months ago I asked chatgpt to search ", "context": "light user, $20-40/month", "use_case": "ML training jobs on GPUs", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}]}, {"label": "Reasonable pricing", "summary": "A light user found the price reasonable at $20 to $40 a month on top of free credits.", "quotes": [{"id": 747584592, "item_id": 46724037, "url": "https://news.ycombinator.com/item?id=46724037", "by": "dataviz1000", "date": "2026-01-22", "stance": "recommend", "quote": "I recommend the service but like I said I haven't tried any others.", "excerpt": "Hey, I use Modal! I don't know how it compares to any other service because I used it once and it worked perfect at price point I found reasonable. Also, working with Claude code and the Modal cli API is perfect for managing all the jobs and moving assets around. I tried writing scripts to manager assets across my local device and Google's colab but it was still not ideal. So two months ago I asked chatgpt to search ", "context": "light user, $20-40/month", "use_case": "ML training jobs on GPUs", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}]}], "caveats": [{"label": "Docker in sandboxes", "summary": "Its sandboxes use gVisor rather than Firecracker, which makes running Docker inside them harder.", "quotes": [{"id": 786594898, "item_id": 49162181, "url": "https://news.ycombinator.com/item?id=49162181", "by": "dregitsky", "date": "2026-08-03", "stance": "compare", "quote": "Modal, but they use gVisor and not Firecracker, which means it's harder for users to run docker inside the box", "excerpt": "For boxes.dev we use E2B - they had everything we needed at the time. I'd love a standardized API / marketplace, but more just to try out new providers with differentiated features or pricing - we probably wouldn't be switching often / using multiple (unless customers asked us for more options). There are so many sandbox providers popping up! Our use case is pretty specific though - one persistent machine that you se", "context": "builds boxes.dev", "use_case": "sandbox VMs for agents", "switched_from": "", "story": "Devtools must be open source"}]}, {"label": "Business sustainability", "summary": "One commenter doubts GPU-as-a-service middlemen reselling public clouds can be profitable long term.", "quotes": [{"id": 747583504, "item_id": 46723969, "url": "https://news.ycombinator.com/item?id=46723969", "by": "bflesch", "date": "2026-01-22", "stance": "warn", "quote": "business model of GPU-as-a-service middleman companies like modal deeply unsustainable", "excerpt": "In his newsletter Ed Zitron hammered down the point that GPUs depreciate quickly, but these kind of reliability issues are shocking to read. The GPUs are so common to fail that they hang out in a 24/7 slack channel with customers like Meta (who apparently can't set up a cluster themselves..). Ed Zitron also called out the business model of GPU-as-a-service middleman companies like modal deeply unsustainable, and I al", "context": "", "use_case": "GPU cloud", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}]}], "who": [{"by": "qkwrv", "context": "enterprise software developer", "url": "https://news.ycombinator.com/item?id=49114019", "date": "2026-07-30"}, {"by": "RichardOdds", "context": "solo founder", "url": "https://news.ycombinator.com/item?id=47021167", "date": "2026-02-14"}, {"by": "dataviz1000", "context": "light user, $20-40/month", "url": "https://news.ycombinator.com/item?id=46724037", "date": "2026-01-22"}], "tags": ["for GPU workloads", "for ML training", "for LLM inference", "for AI agent sandboxes", "for light users"], "use_cases": ["provisioning ml training instances", "inference hosting", "sandbox vms for agents", "spinning up vms for simulations", "hosted sandboxes", "gpu faas workloads"], "compared": [{"slug": "e2b", "name": "E2B", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "claude-code", "name": "Claude Code", "co": 3, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=46669001", "title": "Keeping 20k GPUs Healthy", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49508982", "title": "Apple caught off guard by AI demand for Mac Mini and Mac Studio", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49261444", "title": "item 49261444", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49156111", "title": "Devtools must be open source", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49113999", "title": "Show HN: NegativeEV \u2013 I built a tool to help people see how bad their bets are", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48642510", "title": "Run isolated sandboxes with full lifecycle control: AWS introduces MicroVMs", "mentions": 1}], "sentiment_by_month": {"2025-11": {"compare": 1, "recommend": 1}, "2025-12": {}, "2026-01": {"warn": 1, "recommend": 1}, "2026-02": {"neutral": 1, "recommend": 1}, "2026-03": {}, "2026-04": {}, "2026-05": {}, "2026-06": {"recommend": 1}, "2026-07": {"recommend": 1}, "2026-08": {"compare": 1, "neutral": 1, "recommend": 1}, "2026-09": {}, "2026-10": {}}, "quotes": [{"id": 792237824, "item_id": 49514864, "url": "https://news.ycombinator.com/item?id=49514864", "by": "ronfriedhaber", "date": "2026-08-31", "stance": "recommend", "quote": "Modal significantly improves this. Highly recommend.", "excerpt": "> Just provisioning a new instance and copying the relevant checkpoints and things can take 25 minutes. Modal significantly improves this. Highly recommend.", "context": "", "use_case": "provisioning ML training instances", "switched_from": "", "story": "Apple caught off guard by AI demand for Mac Mini and Mac Studio"}, {"id": 788183125, "item_id": 49261445, "url": "https://news.ycombinator.com/item?id=49261445", "by": "yummyelephant8", "date": "2026-08-11", "stance": "neutral", "quote": "hooked it up to vLLM on Modal", "excerpt": "Like you, I use coding agents like Claude Code, Codex, and Cursor every day. But the other day, I was curious about how Cursor trained their original tab-completion model. So I started wondering: could you post-train an open-source LLM to autocomplete ComfyUI workflows given your current workflow, similar to how Composer autocompletes code? More specifically: Could I post-train an LLM to predict up to four nodes in t", "context": "", "use_case": "inference hosting", "switched_from": "", "story": null}, {"id": 786594898, "item_id": 49162181, "url": "https://news.ycombinator.com/item?id=49162181", "by": "dregitsky", "date": "2026-08-03", "stance": "compare", "quote": "Modal, but they use gVisor and not Firecracker, which means it's harder for users to run docker inside the box", "excerpt": "For boxes.dev we use E2B - they had everything we needed at the time. I'd love a standardized API / marketplace, but more just to try out new providers with differentiated features or pricing - we probably wouldn't be switching often / using multiple (unless customers asked us for more options). There are so many sandbox providers popping up! Our use case is pretty specific though - one persistent machine that you se", "context": "builds boxes.dev", "use_case": "sandbox VMs for agents", "switched_from": "", "story": "Devtools must be open source"}, {"id": 785824305, "item_id": 49114019, "url": "https://news.ycombinator.com/item?id=49114019", "by": "qkwrv", "date": "2026-07-30", "stance": "recommend", "quote": "Modal is awesome, btw", "excerpt": "I know I say it in the title, but I want to be very clear up front: the predominant output of this tool is \"don't place this bet.\" Betting apps want you to bet. If they sense you are tired of losing and might leave, they'll entice you with deposit bonuses, boosted odds and protected picks. There are even people, you see them self-promoting all over X, who sell access to \"good\" bets. Obviously, they don't refund you i", "context": "enterprise software developer", "use_case": "spinning up VMs for simulations", "switched_from": "", "story": "Show HN: NegativeEV \u2013 I built a tool to help people see how bad their bets are"}, {"id": 779023829, "item_id": 48688989, "url": "https://news.ycombinator.com/item?id=48688989", "by": "dbmikus", "date": "2026-06-26", "stance": "recommend", "quote": "Daytona, E2B, OpenComputer, Freestyle, Blaxel, Vercel, Modal, Cloudflare, Tensorlake, Superserve, etc. etc.", "excerpt": "Why do you want to self-host vs. using one of the many providers out there? Daytona, E2B, OpenComputer, Freestyle, Blaxel, Vercel, Modal, Cloudflare, Tensorlake, Superserve, etc. etc. Some of them work by pre-purchasing credits, so you can control the blast radius of spend. Also, if you want a more embedded sandbox runtime as a library instead of a daemon + REST API, you can check out libkrun (and friendly layers on ", "context": "", "use_case": "hosted sandboxes", "switched_from": "", "story": "Run isolated sandboxes with full lifecycle control: AWS introduces MicroVMs"}, {"id": 753368656, "item_id": 47085541, "url": "https://news.ycombinator.com/item?id=47085541", "by": "isoprophlex", "date": "2026-02-20", "stance": "recommend", "quote": "I love modal. I think they got FaaS for GPU exactly right", "excerpt": "> There are no great FaaS options for running GPU workloads I love modal. I think they got FaaS for GPU exactly right, both in terms of their SDK and the abstractions/infra they provide.", "context": "", "use_case": "GPU FaaS workloads", "switched_from": "", "story": "Almost Every infrastructure decision I endorse or regret after 4 years"}, {"id": 752338679, "item_id": 47021167, "url": "https://news.ycombinator.com/item?id=47021167", "by": "RichardOdds", "date": "2026-02-14", "stance": "neutral", "quote": "across Replicate, Modal, and custom GPU clusters", "excerpt": "Hi HN, I\u2019m the solo founder of Snaprookies. I was tired of the \u2018SaaS sprawl\u2019 in my workflow - paying for 5 different AI subscriptions just to get one project done. I built SnapRookies as a unified orchestration layer that routes tasks to specialized models based on the specific creative intent. The Architecture: Rather than a simple wrapper, we use an ensemble approach. For example, our 'Professional Headshot' pipeli", "context": "solo founder", "use_case": "model hosting", "switched_from": "", "story": "Show HN: Snaprookies a unified orchestrated layer for 27 generative AI workflow"}, {"id": 747584592, "item_id": 46724037, "url": "https://news.ycombinator.com/item?id=46724037", "by": "dataviz1000", "date": "2026-01-22", "stance": "recommend", "quote": "I recommend the service but like I said I haven't tried any others.", "excerpt": "Hey, I use Modal! I don't know how it compares to any other service because I used it once and it worked perfect at price point I found reasonable. Also, working with Claude code and the Modal cli API is perfect for managing all the jobs and moving assets around. I tried writing scripts to manager assets across my local device and Google's colab but it was still not ideal. So two months ago I asked chatgpt to search ", "context": "light user, $20-40/month", "use_case": "ML training jobs on GPUs", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}, {"id": 747583504, "item_id": 46723969, "url": "https://news.ycombinator.com/item?id=46723969", "by": "bflesch", "date": "2026-01-22", "stance": "warn", "quote": "business model of GPU-as-a-service middleman companies like modal deeply unsustainable", "excerpt": "In his newsletter Ed Zitron hammered down the point that GPUs depreciate quickly, but these kind of reliability issues are shocking to read. The GPUs are so common to fail that they hang out in a 24/7 slack channel with customers like Meta (who apparently can't set up a cluster themselves..). Ed Zitron also called out the business model of GPU-as-a-service middleman companies like modal deeply unsustainable, and I al", "context": "", "use_case": "GPU cloud", "switched_from": "", "story": "Keeping 20k GPUs Healthy"}, {"id": 736237299, "item_id": 46014831, "url": "https://news.ycombinator.com/item?id=46014831", "by": "mritchie712", "date": "2025-11-22", "stance": "recommend", "quote": "We use e2b.dev, but Modal is great too.", "excerpt": "Some things we've[0] learned on agent design: 1. If your agent needs to write a lot of code, it's really hard to beat Claude Code (cc) / Agent SDK. We've tried many approaches and frameworks over the past 2 years (e.g. PydanticAI), but using cc is the first that has felt magic. 2. Vendor lock-in is a risk, but the bigger risk is having an agent that is less capable then what a user gets out of chatgpt because you're ", "context": "", "use_case": "agent sandbox computer", "switched_from": "", "story": "Agent design is still hard"}, {"id": 735328608, "item_id": 45958038, "url": "https://news.ycombinator.com/item?id=45958038", "by": "TheTaytay", "date": "2025-11-17", "stance": "compare", "quote": "I wonder how it is different from Modal", "excerpt": "This _sounds_ really neat, and I'm very curious about this! If it helps, I wanted to give you my stream-of-consciousness as I explored your post and site as a _total_ newcomer. I am a bit confused about what it can dop, and I bet it's really easy to make me unconfused through some site tweaks, and you seem eager for feedback: My thoughts: \"Woah, this sounds neat. I wonder how it is different from Modal...I wonder if ", "context": "", "use_case": "batch execution", "switched_from": "", "story": "Show HN: Bsub.io \u2013 zero-setup batch execution for command-line tools"}]}