Rate limits

Rate limits cap how much work a single account can push through the condense proxy in a window of time. condense applies two independent limits, a request rate and a compaction credit pool, and you hit whichever runs out first.

Why rate limits exist

Like any shared API, condense meters usage so that one account can't degrade the service for everyone else. Limits protect the GPU-served compaction backends from overload, keep latency predictable, and bound the cost of a runaway client. They apply at the account level (and, for teams, the organization level), not per key.

How they work

Two limits govern every request:

Limit	Unit	What happens at the edge
Request rate	Requests per minute, a per-account limit that scales with your plan, plus a flat per-IP catch-all	Excess requests get `429 Too Many Requests` with a `Retry-After` header. Retry after the delay.
Compaction credits	Credits per month, drawn per metered compaction	When the pool is exhausted, requests pass through uncompacted rather than failing, you keep talking to your provider, just without the token savings.

Your provider's limits still apply. condense caps the proxy edge; the upstream model call is separately subject to your own provider account's rate and spend limits.

Limits by plan

The per-minute request rate scales with your plan; the monthly compaction credit pool scales too. Your account's live numbers are in your dashboard.

Plan	Price	Requests / minute	Compaction credits
Free	$0/mo	`20`	Starter pool for trying condense out
Starter	$5/mo	`60`	Individual daily-driver pool
Pro	$20/mo	`120`	Heavy individual pool
Max	$50/mo	`240`	Power-user / small-team pool

Paid credit customers. While we scale capacity, the per-account limit for paid credit customers is currently agreed individually to guarantee availability for your workload. Reach out and we'll set your limit with you directly.

A flat 300 requests/minute per IP also applies as a catch-all, so a single address can't rotate through accounts. Teams and direct-API keys aren't tier-throttled on this path, they run on pooled billing with separate accountability.

Teams share one credit pool across members with per-member budgets. Need a dedicated pool or higher ceilings? Reach out via contact.