Dissent
Frontier models. Clear terms.
Use frontier models through Dissent with a defined allowance and a clear record of usage.
One plan for model access.
Dissent offers managed model access under an agreed utilization plan. Max combines a monthly charge with a weekly API-equivalent credit allowance. The pricing page sets out the offer.
Access is arranged directly. We confirm the start date, model availability, capacity, and any additional charges before you begin.
Agree the plan. Start using it.
- Request access. Tell us what you need to use and how much capacity you expect.
- Confirm the terms. Agree on the available models, allowance, start date, and payment schedule.
- Use your allowance. Access is provided through the arrangement agreed with Dissent.
- Review usage. Compare recorded utilization, API-equivalent value, and your Dissent cost.
Model availability.
Available models and reasoning settings are confirmed for your access arrangement. The usage study on this site recorded GPT-6 Astra and GPT-5.6 Sol activity. Its dollar comparison prices every recorded response at the same Astra Standard benchmark.
Your agreement identifies the models available to you. The public sample is not a complete roster or a guarantee of access to any particular model.
Usage, value, and cost.
Recorded usage
Input and output tokens from retained task records, deduplicated across imports. Cached input is part of input; reasoning tokens are part of output.
API-equivalent value
What the recorded token mix would cost at the stated API benchmark, including applicable cache and context-length rates.
Your Dissent cost
Your agreed charge for the same service period. It is separate from the API benchmark and your credit allowance.
Savings with Dissent is the difference between API-equivalent value and your Dissent cost for the same period, when both are known. Unused credits are not savings. Missing usage is left unreported until reconciled.
About the public usage study
The measured sample covers September 6, 2026 at 09:48 EDT through September 13, 2026 at 09:48 EDT: 7,881,719,790 tokens across 55,010 unique responses. It is an operational sample, not a claim about any individual client.
The 30-day and YTD views are planning scenarios extrapolated from that sample under a modeled operating scale. YTD means January 1 through September 13, 2026, inclusive. Neither projection is a historical usage counter, a capacity guarantee, a customer bill, or realized savings.
API-equivalent values use published Astra Standard pricing verified September 13, 2026. The measured week's cache-aware benchmark is $11,194.43. Public totals contain no customer identities or private access details.
Keeping the record current
Timestamped captures preserve progress between imports. Month-end reconciliation rechecks the full calendar month and removes duplicates. Reports distinguish retained records from missing coverage; quota resets do not by themselves establish a complete usage history.
A record you can inspect.
Reports separate token usage, its API-equivalent value, and your agreed charge. Only usage assigned to your account belongs in your report. A report or review does not itself authorize a deployment, purchase, or other external action.
One key for the right model.
Standard Dissent accounts can use a single API key. The router classifies the task, selects an approved model and reasoning profile, and keeps the route in the workspace record. Read the API guide to see the request shape.
Access by arrangement.
Max pricing is described on the pricing page. Invoices are paid manually. There is no public checkout.
The client portal surfaces are live in private beta. Email authentication, hosted usage ingestion, Stripe invoice links, and automated provider execution are still being connected. The public usage study is a dated snapshot; it is not a live customer usage ledger.
Dissent's automated coordination and MCP work remains separate development work. It is not a production integration offered by this page.