SF Compute
Compute markets · Series A
A market for GPU time: reserve a cluster by start date and length, resell unused hours.
- Makes
- Compute markets
- Stage
- Series A
- Open roles
- 9
- First look
- Not yet
The startups we follow closely. We use each product, then film what we find. No film is out yet.
14 of 14 startups
Compute markets · Series A
A market for GPU time: reserve a cluster by start date and length, resell unused hours.
Post-training · Seed
Free open-source software to fine-tune and run open models on your own machine.
Computer use · Series A
Isolated sandboxes an agent can start in milliseconds to run the code it wrote.
Inference
Hosted open-model inference at realtime, async and overnight-batch prices.
Computer use · Series A
A fresh microVM per agent session, so generated code runs away from your servers.
Local & on-device AI
Links the computers you own into one machine big enough for a large model.
Foundation models
Diffusion language models that write a whole answer at once, sold by API.
Local & on-device AI
A desktop app that downloads and runs open models on your machine.
Local & on-device AI · Series B
The one-command way to run open models on your own computer.
GPU compute · Series A
GPU clusters plus reinforcement-learning tools to train a model on your own work.
Agent frameworks · Series A
The Python data-checking library, now an agent framework, evals and tracing.
Foundation models · Series B
A Tokyo lab selling routed multi-model access and Japanese-language models.
Durable workflows · Series A
Runs long background jobs and agent loops in TypeScript with no timeout.
GPU compute
A marketplace for renting other people’s spare GPUs by the hour.
How a startup gets here. We use the product ourselves. Security and privacy problems are reported to the team privately and never published. A first look carries no score.
Is this your company? Claim its page to answer readers and list your roles yourself. Claim your page