Ollama Cloud pricing
Ollama Cloud's paid plans include a fixed amount of usage credits each month, spent in tokens at each model's published per-million-token rate. The $0 Pay as you go plan has no subscription and pays per request with usage credits. The information below reflects the plans shown at ollama.com/pricing, re-read automatically every day and last verified .
Plan comparison
- Run models locally
- No subscription
- 3 concurrent requests
- No service fees
- 3 concurrent requests
- $60 of usage credits per month
- Access to larger pro models
- Usage alert at 90% of included credits
- 10 concurrent requests
- $300 of usage credits per month
- Early access to the newest models
- Everything in Pro
- Unlimited users
- $1,000 of usage credits per month, shared across the team
- 10 concurrent requests
- Centralized billing and administration
- Priority support
- Everything in Team
- Model access controls and per-key cost budgets
- Private Slack channel with dedicated support
- Custom security questionnaires
How Ollama Cloud billing works
Usage is measured in tokens at each model's rate — separate prices per million input, cached-input and output tokens, listed in the table below. On paid plans each request is priced at those rates and drawn from the included credits first, then from any purchased balance. Pay as you go pays per request with usage credits and includes no monthly credits.
- Included credits on Pro, Max and Team reset monthly on the same day of the month your subscription started, including on annual plans.
- No rollover. Unused included usage is replaced at each reset, not carried forward.
- Extra usage on every plan. Pay as you go, Pro, Max and Team can all buy usage credits; on paid plans included plan credits are spent first, then the purchased balance. Team usage draws from one balance shared by the organization.
- Alert at 90%. Paid plans get an email at 90% of the included monthly usage; it can be turned off in settings.
Ollama Cloud model pricing
Prices are in USD per million tokens, as published. The last column is a computed output-only ceiling — how many output tokens $60 of Pro credits would buy if every token were output at that model's rate. Real requests also spend input and cached-input tokens, so treat it as an upper bound for comparing models, not a quota.
| Model | Input | Cached input | Output | ≈ Output tokens per $60 |
|---|---|---|---|---|
| beam | $0.125 | $0.015 | $0.4 | 150M |
| clef | $0.24 | — | — | — |
| deepseek-v4.1-flash | $0.15 | $0.003 | $0.6 | 100M |
| deepseek-v4-pro | $0.66 | $0.022 | $1.98 | 30M |
| gemma4 | $0.14 | $0.05 | $0.4 | 150M |
| glm-5.3 | $1.4 | $0.26 | $4.4 | 14M |
| glm-5.3-flash | $0.15 | $0.03 | $0.5 | 120M |
| glm-5.2 | $1.4 | $0.26 | $4.4 | 14M |
| gpt-oss:120b | $0.15 | $0.014 | $0.6 | 100M |
| gpt-oss:20b | $0.07 | $0.035 | $0.3 | 200M |
| kimi-k3.1-pro | $2 | $0.2 | $10 | 6.0M |
| kimi-k3 | $3 | $0.3 | $15 | 4.0M |
| kimi-k2.7-code | $0.95 | $0.19 | $4 | 15M |
| kimi-k2.6 | $0.95 | $0.16 | $4 | 15M |
| mimo-v2.6-flash | $0.14 | $0.0028 | $0.28 | 214M |
| mimo-v2.6-pro | $0.435 | $0.0036 | $0.87 | 69M |
| minimax-m3.1-flash | $0.15 | $0.015 | $0.6 | 100M |
| minimax-m3 | $0.6 | $0.12 | $2.4 | 25M |
| minimax-m2.7 | $0.3 | $0.06 | $1.2 | 50M |
| mistral-large-4 | $0.7 | $0.07 | $2.1 | 29M |
| mistral-large-3 | $0.5 | — | $1.5 | 40M |
| muse-spark-open | $1.25 | $0.15 | $4.25 | 14M |
| nemotron-3-nano | $0.06 | — | $0.24 | 250M |
| nemotron-3-super | $0.015 | $0.015 | $0.6 | 100M |
| nemotron-3-ultra | $0.1 | $0.1 | $3 | 20M |
— means no rate is published for that tier; it is not zero. Clef is an input-only decision model, returning one JSON response through the SystemOne API without streaming or autoregressive text generation. Its output-token ceiling is unavailable; decision usage is not comparable to text-generation TPS.
What is Ollama's peak pricing?
A few models bill at a second, higher rate inside a published window. Peak pricing applies between 12:00 and 18:00 UTC, Monday to Friday. Models absent from this table bill at the standard rate around the clock.
| Model | Peak input | Peak cached input | Peak output |
|---|---|---|---|
| deepseek-v4.1-flash | $0.3 | $0.006 | $1.2 |
| deepseek-v4-pro | $1.32 | $0.044 | $3.96 |
Concurrency limits
The number of cloud requests you can have in flight simultaneously is plan-gated:
| Plan | Concurrent requests |
|---|---|
| Pay as you go | 3 |
| Pro | 3 |
| Max | 10 |
| Team | 10 |
Requests beyond the concurrency limit are queued. When the queue fills, new requests are rejected until a slot opens.
Still on an older Pro or Max plan?
Pro and Max subscriptions started before the credits-based plans keep their original terms — a usage allowance with a session limit that resets every 5 hours and a weekly limit that resets every 7 days — until the subscriber switches. Switching grants the new plan's full monthly credits immediately and removes the session and weekly windows; the reset date stays on the original subscription date.
How these prices affect benchmark results
The benchmarks on this site run on a Pro subscription — a legacy one, taken out before the credits-based plans, which is why cadence on the Pro rows is not budget-constrained. Speed numbers reflect Pro-tier infrastructure. Pay as you go users may see different throughput under load.
See the methodology page for the full measurement spec.
Weighing a monthly credit plan against per-token billing through a router? Ollama Cloud vs OpenRouter covers where each one wins, and is explicit about which side of it we actually measure.
Frequently asked questions
How much does Ollama Cloud cost?
Ollama Cloud's $0 entry plan is Pay as you go: no subscription, 3 concurrent requests, and usage paid per request with usage credits. Pro is $20/mo or $200/yr and includes $60 of usage credits per month; Max is $100/mo with $300 of credits; Team is $500/mo with $1,000 of shared credits and unlimited users; Enterprise is custom pricing. Every plan, including Pay as you go, can buy usage credits.
Does Ollama Cloud charge per token?
Yes. Usage is measured in tokens at each model's published rate per million input, cached-input and output tokens. Paid plans draw it from their included credits; Pay as you go pays per request with usage credits — so a heavier model spends credits faster than a light one for the same token count. For example, beam is $0.125 per million input tokens and $0.4 per million output tokens. Running Ollama locally on your own hardware remains free and unlimited.
What is the difference between Ollama Cloud Pay as you go, Pro and Max?
Credits and concurrency. Pay as you go has no subscription fee and runs 3 concurrent requests; usage is paid per request with usage credits. Pro runs 3 concurrent requests with $60 of monthly credits and access to the larger pro models. Max runs 10 concurrent requests with $300 of monthly credits and early access to the newest models.
Do unused Ollama Cloud credits roll over?
No. Included credits refresh monthly on the same day of the month your subscription started, including on annual plans; whatever is unused is replaced, not carried forward. Ollama emails paid accounts at 90% of the included monthly usage.
What is Ollama's peak pricing?
Peak pricing applies between 12:00 and 18:00 UTC, Monday to Friday. Inside that window the affected models bill at a higher rate: deepseek-v4.1-flash at $0.3 per million input and $1.2 per million output tokens, deepseek-v4-pro at $1.32 per million input and $3.96 per million output tokens. Every other model bills at its standard rate around the clock.