Pinned Tweet
Today we're announcing that Modal has crossed $1B in annualized run rate, as well as raised a $750M Series D at a $15.75B valuation, led by @Accel.
Since our Series C, usage has tripled and we’ve grown more than 15x year-over-year.
We are deeply grateful to the developers and companies building on Modal. There's still so much more to build, and we can't think of a better time to be building software.
gVisor is joining the @CloudNativeFdn, and Modal is becoming one of its maintainers.
We’ve run gVisor on Modal since day one. Along the way we've upstreamed: GPU support, multi-node training, snapshots, and more. All of it open-source.
We’re incredibly proud to grow from being a team that depends on gVisor to a team that also gets to help steer it.
How does a large, global company like @DoorDash enable its product teams to build with AI quickly, but also safely and securely?
DoorDash’s GenAI platform team helps engineers across the company build products securely with AI, while leaving room to tinker and experiment. They get access to production-ready model APIs and can run batch jobs, online apps, and agents, while still maintaining the level of governance, performance, and security DoorDash requires.
Hear from Skyler Roh, AI Platform Engineer at DoorDash, on how they built it.
youtu.be/vbR24n6VI3w
7
878
Compute shouldn’t be the bottleneck for your next big idea.
We're launching Modal for Builders, a program that provides compute credits to builders working on ambitious AI projects.
Have a demo you’ve been wanting to build? A research question you want to explore? Or another idea that needs compute?
We’d like to hear about it. Apply with your idea and an estimate of the compute you’ll need. Projects may also be featured on Modal’s social channels.
Some projects we’ve backed in the past:
GPT-2 from scratch by @GreenCodeCodes: a YouTube series on training an LLM from scratch
xitter.twitterapi.xyz/GreenCodeCodes/status/…
I'm launching a series where I build my own LLM from scratch to raise money for cancer research :)
Huge thanks to @modal and @real_deep_ml for sponsoring!
Any other companies interested? All the sponsorship money goes to charity.
Vmax is building Campaign, a platform for RL task generation using self-play.
We worked with them to teach agents to use the Modal SDK, translating Python code into working Modal workloads by generating tasks that target their failure modes.
Modal's elastic compute lets them run and check solutions, from GPU calculations to inference endpoints that reuse state across requests.
1/ As a teaser for Campaign, Vmax’s infrastructure for automated task generation, we are sharing work we did earlier this year for @modal using self-play to generate tasks that target weaknesses of models when using the Modal-SDK.
Frontier AI data lab @SnorkelAI uses Modal Sandboxes and Modal Inference to power millions of agent evals each month.
Snorkel uses evals to test agents on rigorous and diverse benchmarks ranging from coding to infrastructure eng through data science.
Modal lets them flexibly serve open-weight models and use a mix of CPU and GPU Sandboxes at scale, cutting their sandbox bill by 20%.
A foundational text of algebraic geometry, now formalized in Lean.
Congratulations to the @axiommathai team!
Some moments from Runtime we're playing on a loop.
Want to rewatch your favorite talk? Sessions are live at modal.com/runtime.
1,300+ square miles. 172 H100-hours. An isometric map of the SF Bay Area that you can glide through, search, and share by @jus10chavez.
Powered by Modal.
Introducing Isometric SF A pixel-art map of the entire SF Bay Area, over 1,300 square miles, all painted by Klein, a 4B-parameter image-to-image model, on @modal.
Explore the map now!
isometricsf.com
Kimi K3.1 Pro from @Kimi_Moonshot is now available on Modal.
A major post-training update to Kimi K3, with expanded reinforcement learning for stronger coding, financial research, and long-horizon agentic work.
Kimi K3.1 Pro supports text and images with 1M context.
We collaborated on the model bringup with @sgl_project, validating checkpoint accuracy and optimizing the serving path to support Kimi K3.1 Pro on Day 0.
xitter.twitterapi.xyz/sgl_project/status/211…
Day 0 support for Kimi K3.1 Pro from @Kimi_Moonshot coming this week on Modal.
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
In July, @muni_bio and @onepot_ai co-hosted Ebolathon to find small molecules that could block Bundibugyo ebolavirus from entering cells without harming the cells themselves.
Teams designed 140 molecules in just one Saturday in San Francisco. All were synthesized in 11 lab days, followed by 22 lab days of testing. The results are in: 26 of the 140 molecules were preliminary hits—an 18.6% hit rate.
In part, powered by Modal.
Binding is nice, but it’s getting boring. If you’re not thinking about biological function, you’re ngmi.
At our last @muni_bio hackathon co-hosted with @onepot_ai, we set one ambitious goal: neutralize the latest threat of Ebola using small molecules.
The wet lab data is in, and we’re excited to announce that teams were successful, including our agent that repurposed... an anti-itch gel?
muni.bio/research/hits-from-…
Read more about the hackathon: muni.bio/research/hits-from-…
3
1,450
And we added first-class support for post-training Beam yourself.
Get started with Modal Dojo now: dojo.modal.dev/
Reflection Beam is now available on Modal.
Beam is a 501B-parameter open-weight MoE (23B active) from @reflection_ai built for coding, agentic workflows, and reasoning. It supports a 1M token context window, with token efficiency a core focus.
In early access, the Oura Health Intelligence team found that "Beam had fewer context leaks and was much more grounded in its answers than other frontier models we evaluated.”
Nova got Opus 4.8 to edge past Fable 5 on Terminal-Bench 2.1.
Modal Sandboxes let the team decouple Nova’s agent loop and automatic review from tool execution, with fast startup and changes deployed in seconds to keep iteration quick.
The result: 83.9%, up five percentage points from Opus 4.8’s Claude Code baseline and just ahead of Fable 5’s 83.8%.
We made Opus 4.8 perform like Fable 5.
Same model. No fine-tuning.
Terminal-Bench 2.1:
Opus 4.8 in Claude Code: 78.9%
Opus 4.8 in Nova Agent OS: 83.9%
Fable 5: 83.8%
Built together with our friends at @modal. The frontier isn't only the model anymore.
novaagentos.ai/news/how-nova…
Students shouldn’t have to scale down their research ideas because they can’t get access to compute.
The @generalcatalyst Research Fellowship will support 20 undergraduates with up to $25K in compute from Modal to power their most ambitious ideas.
We're excited to see what you build!
Introducing @generalcatalyst Research Fellows - a program for undergraduates ready to pursue their most ambitious ideas.
We’re supporting approximately 20 exceptional undergraduate builders from leading schools. In partnership with @Modal, each fellow will receive up to $25K in compute for a 90-day project of their own.
For the bold, the curious, and those building what they wish existed: applications are open now. Tell us what you want to build! generalcatalyst.com/research…
We recently announced that Modal has crossed $1B in annualized run rate, as well as raised a $750M Series D at a $15.75B valuation.
Our CEO and co-founder @bernhardsson stopped by @NYSE to share more about what's being built on Modal today and what's next for the company post-fundraise...
Check out the full interview: tv.nyse.com/videos/modal-ceo…
1
1
1,012
DeepSeek v4.1 Flash 1017 from @deepseek_ai is now available on Modal. An update of DeepSeek v4.1 Flash, it improves substantially on reasoning, coding, agentic tasks, and visual understanding.
Learn more and get started: modal.com/library/deepseek/d…
4
1,068
.@markovrobotics is teaching robots to tackle complex manipulation tasks with fewer than ten hours of robot data. Their bet: video already contains most of the physical reality a robot needs to act.
They use Modal for video processing, world model training, and remote inference.
Using Modal Functions, Markov fans out preprocessing across thousands of CPU and GPU containers, reaching upwards of 500 GPUs at once on the heaviest stages.
Work that could take a full day on traditional setups completes in 20-30 minutes on Modal, thanks to the ability to parallelize work over on-demand compute.
Read the full post: modal.com/blog/markov-case-s…
1
2
987
.@WeAreLegora runs 2,000 agent evals in parallel on Modal Sandboxes, testing a legal agent that works for hours across thousands of documents.
Every run boots the entire application, pinned to a specific commit, so engineers can measure the impact of their changes in a realistic environment.
Legora chose Modal for scalability, low startup latency, and the flexibility to run its large, complex app. With Modal providing reproducible environments at scale, Legora’s agents can now create cases, launch experiments, and analyze results autonomously.
With engineers managing the process, Legora used this approach to cut agent cost by 30% through smarter compaction and custom tool search.
Read the full post: modal.com/blog/how-legora-ru…
5
937
We worked with @SakanaAILabs to optimize their Sakana Namazu API down to the kernel level:
→ Selective NVFP4 quantization (23k+ experts)
→ DFlash draft model trained on raw token streams (+25% acceptance)
→ In-house prefix KV routing + serverless autoscaling
Read the full deep dive: modal.com/blog/sakana-case-s…
