Pricing

Delphina is priced on two principles:

1. Pay for what you use. Billing tracks the work the agent does.
2. Price aligned with value and cost of work. Compute and inference are metered separately so cost scales with what the agent actually consumed on your behalf.

​

Tiers

Enterprise pricing can be seat- or consumption-based. See delphina.ai/pricing for what’s included in each tier. The rest of this page covers how consumption-based pricing is metered.

​

Credits

Credits are consumed as the agent does work. Two things consume credits:

• Compute — time the agent spends running on Delphina infrastructure
• Inference — tokens sent to and received from the underlying LLMs

Org admins can see usage on the Usage page.

​

Compute consumption

Compute is billed by wall-clock time the agent is actively running.

​

Inference consumption

Inference is metered per token, in credits per 1M tokens. Inference has two components: provider token costs and Delphina token costs. Your inference bill is the sum of the two.

​

Provider token costs

Provider token costs are converted into credits based on public provider rates, at the credit exchange rate of $3 per 100 credits, with no markup. See current rates: Anthropic, OpenAI, Google Gemini.

​

Delphina token costs

Delphina token costs are model-agnostic — they apply the same way whether the agent is using Claude, GPT, Gemini, or any other supported model — and vary by mode type. You select the mode when starting a chat or workflow:

• Knowledge — Agent calls that build and maintain your context layer.
• Analytics — Day-to-day chat and workflow runs.
• Research — Long-horizon, multi-step research that explores many alternatives.

​

Terms

• Billing cycle — all usage, including overages on included monthly credits, is billed monthly in arrears.
• Credit rollover — included monthly credits reset each month and do not roll over.
• Pre-paid commitments — discounts available for committed credit purchases. Contact your account team.