Plans that scale with you
Choose a plan that fits your needs, and scale up as your usage grows.
Serverless Inference
Plans include pre-paid tokens at a discount compared to list price. Any usage above the included amount is charged on top. Any remaining balance is rolled over to the following month.
Trial
€0 / month
For the hobbyist
€5 free starting credits
Basic rate limits*
Community support
No monthly cost
Starter
€25 / month
For the small team
€25 monthly top-up
Increased rate limits*
Email support
Automatic billing
Team
€50 / month
For developer teams
€55 monthly top-up
Higher rate limits*
Priority support
10% extra credits
Enterprise
€500 / month
When you need scale
€550 monthly top-up
Maximum rate limits*
Dedicated support
10% extra credits
Invoice and Peppol billing
* Rate limits are shared per account.
Berget Code
Seat-based pricing for a predictable cost for your team. Choose the number of seats and assign them to your team members.
Standard
€150 / month per seat
For developer teams who want frontier-level capabilities for agentic coding
No token charge
Access to all frontier-level models
Standard rate limits
OpenCode and Pi integrations
Support
Pro
€300 / month per seat
For power users who need more headroom
Everything in Standard
Increased rate limits
Higher concurrency
Priority support
* Rate limits are shared per seat.
Model pricing
Frequently asked questions
What's the difference between Standard, Pro, and Summit for Berget Code?
All Berget Code plans are seat-based with a fixed monthly cost — you never pay per token. The difference lies in the size of your token quota per period and how many agents you can run in parallel. Standard suits most users, Pro is for heavier agentic workflows, and Summit is for teams running at scale.
You can upgrade at any time in the console; downgrades take effect from your next billing cycle.
How do quotas work in Berget Code?
Quotas reset across three windows: daily, weekly, and monthly. Each window starts with the first request after the previous window ends — so you can run intensive sessions without exhausting your monthly budget in the first week.
Frontier models like Kimi K3 consume quota faster: once you pass 50% of your total quota, Kimi K3 is rate-limited, and you'll get the most value from your plan by switching to a cheaper model such as GLM-5.3-Flash. You can see your daily, weekly, and monthly consumption in the console under Berget Code.
Do I need a monthly plan, or can I pay per use?
The Trial plan includes €5 in starting credits so you can try the API without a subscription — a card is required at signup to prevent abuse, but nothing is charged. After that, an active plan (from €25 / month in prepaid credits) is required for continued API access — we don't offer pure pay-as-you-go without a plan.
On a paid plan you can add one-off top-ups or set up automatic top-up rules. Unused credits roll over to the next month and remain valid for 12 months.
What happens to my credits if I cancel?
You can keep using your remaining credits for 30 days after cancelling. If you reactivate within that window, you keep all rolled-over credits — otherwise they expire. Prepaid credits are non-refundable.
This applies to Serverless Inference credits — Berget Code seats are flat-rate with no credits to carry over.
Are prices including or excluding VAT?
All listed prices exclude VAT. Swedish consumers are charged 25% VAT; EU businesses with a valid VAT ID are reverse-charged (0%); non-EU customers pay no VAT. A €50 plan therefore shows as €62.50 for a Swedish private individual.
Where do I find my invoices and receipts?
Invoices are emailed automatically with each payment. A self-serve invoice view in the console is in development — until then, contact support and we'll resend any invoice.
Company name and VAT ID on invoices can currently only be changed by support.
How are cached (prompt cache) tokens billed?
Cached input tokens are currently billed at the same rate as regular input tokens, and cache usage isn't broken out separately in reporting yet. Discounted cache-token pricing is a high-priority item on our roadmap.
Keeping the stable part of your prompt first maximises cache hits.