sllm: Share a GPU node with other developers for affordable LLM access

What it does

sllm lets developers rent GPU capacity on a shared node to run large language models. Multiple developers split the cost of a single GPU node, with each reserving a throughput slot. The service runs vLLM and offers an OpenAI-compatible API, so users can swap their base URL without rewriting code. Traffic is not logged, making the setup private by default.

Who it is for

Developers who want to run inference on large models but don't need a dedicated node. The service targets those with moderate throughput needs (15-25 tokens per second mentioned as typical) rather than high-traffic production applications. It appeals to builders who want to experiment with models like DeepSeek V3 without committing to enterprise-grade infrastructure.

Pricing

Prices start at $5 per month for smaller models. The launch post frames the economics around a DeepSeek V3 setup at $14,000 per month if run solo. The site does not show detailed pricing tiers or model-specific rates.

How it stands out

The cohort model is the core differentiator. Developers reserve a spot with a credit card but are not charged until the shared node fills. This means no payment until actual capacity is allocated. Most GPU rental services charge per hour or month regardless of actual utilization. The OpenAI-compatible API removes switching friction for developers already building with that interface. Privacy is explicit: no traffic logging, which some developers prefer over logging-heavy offerings.

What a founder should check

First, validate the actual cohort fill rate and how long developers wait between booking and activation. If cohorts sit half-full for weeks, the economic advantage collapses and users may abandon the service. Second, test the switching cost carefully. While the API is OpenAI-compatible, any differences in error handling, response format, or rate-limiting behavior could force code rewrites and create friction. Third, examine the moat against larger cloud providers adding shared GPU offerings. Compute suppliers like Lambda Labs, Runpod, and major clouds (AWS, GCP, Azure) can easily undercut on price if market traction proves there is demand. The cohort mechanics provide some lock-in once a developer is part of an active group, but that advantage vanishes if nodes are under-utilized or if incumbents copy the model.

Thinking of building something like this?

Every launch here is a competitor to somebody's idea. If yours is close, check it against the market before you build: the Full Check names the rivals, the prices and the gaps.