Layer 2.5 · head to head
Together AI vs RunPod
Tokens against GPU-hours: the crossover calculation almost nobody does.
| Together AI | RunPod | |
|---|---|---|
| Can you leave | Sticky. Leaving means rewriting against a different interface, not changing a hostname. | Movable with effort. Expect to redo images, storage wiring and networking. |
| Counterparty | Low for serverless — you are buying tokens, not capacity. High if you build a product on one provider's fine-tune tooling and its model catalogue. | Low exposure for you: you are renting by the second, so the switching cost of a supplier failing is hours, not quarters. That is the honest advantage of the marketplace model. |
| What it is | You never see a GPU. You send a request and pay per token or per second of execution. | Brokers capacity somebody else owns. Cheapest headline rates, and the machine you get is not the machine you chose. |
| How you reach it | You call an endpoint. There is no infrastructure to see, and no infrastructure to move. | You push a container image and it runs. No cluster to operate. |
| Accelerators | H100, H200, B200 behind the API; dedicated clusters available | H100, H200, A100, L40S, RTX 4090, RTX 5090, and a long consumer tail |
| Regions | US | Global, community and secure clouds |
| Pricing model | Per token for serverless inference; per GPU-hour for dedicated endpoints and training clusters. | Per-second billing, on-demand and spot. Two tiers: 'Secure Cloud' in real datacentres, 'Community Cloud' on other people's hardware. |
| Getting started | API key. | A few dollars. |
| Capacity | Good on popular open-weight models. | Generally good on consumer parts, variable on datacentre parts. |
| Ownership | Private. | Private. |
Together AI
For: Teams serving open-weight models who do not want to operate a GPU at all.
The catch: Per-token pricing hides utilisation. It is cheaper than a dedicated GPU right up until it is dramatically more expensive, and the crossover is a real calculation nobody does before signing up.
Economics: The correct comparison is not against other token prices — it is against your own GPU-hour cost at your actual utilisation. Below roughly 40% duty cycle serverless usually wins; above it, rarely.
RunPod
For: Fine-tuning, batch jobs, inference experiments, anyone whose workload can checkpoint and move.
The catch: Community Cloud is somebody's machine somewhere. For anything with a compliance story attached, that distinction is the whole decision, and it is easy to miss in the pricing table.
Economics: Reported around $2/hr for an H100 on-demand — roughly a third of CoreWeave list. Per-second billing genuinely matters for bursty work.
Neither table row is a price
Deliberately. Published on-demand rates at this layer move weekly, and essentially nobody signing a real contract pays them — every serious buyer pays less than every list figure either of these companies publishes. Quoting one here would date this page within a month.
The GPU rental price index carries dated, sourced figures instead, and the durable finding there is the spread: the identical H100 rents from roughly $1.38 to $12.29 an hour depending only on who you rent it from.
The layers underneath both
Whichever you pick is renting you chips in a building that needs power. In 2026 that is the constraint that binds: Microsoft has disclosed an Azure backlog it cannot fill for want of megawatts rather than accelerators, and the US interconnection queue exceeds 2,600 GW with roughly 80% of projects withdrawing before they energise.
Layer 2 — Silicon · Layer 1 — Energy · The interconnection queue