Ollama pricing guide details
Pricing Overview
Ollama sells two different things under one name, and separating them is the whole pricing question. The runtime is MIT-licensed and free: install it, run models on your own hardware, pay nothing per token or per seat, forever. Ollama Cloud is a metered inference service for models too large to hold locally, priced in monthly usage credits from $0 to $500/mo with an Enterprise tier above that.
So the starting price is $0, and for many buyers it stays there. The decision is not which tier to buy but whether you need the cloud at all — which comes down to two limits on the free plan: the size of model your hardware can hold, and one concurrent request.
Plan Comparison
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Plan | Price | Usage credits | Concurrency | Model access |
|---|---|---|---|---|
| Free | $0 | Starter credits | 1 request | Local models, plus starter cloud models |
| Pro | $20/mo, or $200/yr | $60 per month | 3 requests | Larger pro models |
| Max | $100/mo | $300 per month | 10 requests | Early access to newest models |
| Team | $500/mo | $1,000 per month, shared | 10 requests | Unlimited users, centralised billing, priority support |
| Enterprise | Custom | Negotiated | Negotiated | Model access controls, cost budgets, private support channel |
Annual billing on Pro works out at $16.67/mo, a 17% saving. Note what the credit allowances actually mean: Pro gives $60 of usage for $20, and Max gives $300 for $100. Both are a three-to-one credit-to-dollar ratio, so the paid tiers are not a volume discount curve — they are the same rate with higher ceilings and more concurrency. Team is the exception worth pricing carefully: $500/mo for $1,000 of shared credits is the same ratio again, but it adds unlimited users, which is what makes it a team plan rather than a bigger Max.
Hidden Costs and Considerations
The real cost of the free tier is hardware. Running a capable model locally wants an Apple Silicon machine with substantial unified memory or a discrete GPU, and that capital cost does not appear on any pricing page. Quantisation is the usual workaround, and it trades accuracy for fit.
On the paid tiers, credits are monthly and the pages give no indication that unused credits roll over, so a spiky workload pays for a ceiling it does not use. Concurrency, not credits, is often the binding limit: three simultaneous requests on Pro is low for anything with multiple users, and reaching ten costs $100/mo.
Cost Estimates by Team Size
A solo developer running coding agents on an existing laptop pays $0 — this is the common case, and it is genuinely free rather than a limited trial.
A five-person team that wants larger cloud models occasionally: five Pro seats at $20/mo is $100/mo for $300 of pooled-by-hand credits, but Team at $500/mo gives $1,000 of genuinely shared credits and centralised billing. Below roughly ten people, individual Pro subscriptions are cheaper; above that, Team wins on administration alone.
A twenty-person team is squarely Team territory at $500/mo, or Enterprise if model access controls and cost budgets matter.
How Ollama Pricing Compares
Against per-token providers, Ollama Cloud is not the cheap option. Together AI serves open models from $0.04 per million input tokens up to $3.00 input and $15.00 output per million, and Mistral's La Plateforme also bills per token; both bill only what you use with no monthly floor. Modal, another self-serve compute platform, runs a free Starter tier and $250/mo for Team — half Ollama's Team price, though with a different unit of account.
What Ollama charges for is one API and one vendor across local and hosted inference. If your inference is mostly local with occasional overflow, that coherence is worth the premium. If it is mostly hosted, price the token equivalent first.