By James Miano • 4 min read

Token limits removed for paid accounts

Yesterday paid accounts got 20M tokens per minute and 1B per day. As of today the token counters are off: the only metered numbers are requests per minute and requests per day.

Yesterday morning we raised paid-account ceilings to 20M tokens/minute and 1B tokens/day. By the end of the day the answer was simpler: remove the token caps altogether.

Effective now, for every topped-up account:

  • Tokens per minute: no limit.
  • Tokens per day: no limit.
  • Requests per minute: 3,000 (unchanged).
  • Requests per day: 250,000 (unchanged).

The enforcement engine skipped reading token counters this morning: they were removed from the request path of paid accounts. A 429 on the paid tier now means you crossed a request count, nothing else.

Why drop it

The 1B/day ceiling was defensible when the platform saw a few million tokens a day. The last 48 hours moved that number by two digits: a single account pushed 118.9M tokens in one day, plus burst sessions that flouted minute windows, all inside normal billing. A token ceiling on a pay-per-token product is a sentence with a contradiction in the middle. You are already billed per token: the ceiling only caps growth, never risk.

Requests per minute and per day stay metered. They are the two numbers that actually protect shared capacity from a runaway retry loop, and they are the ones upstream abuse would hit first.

What does not change

Free-tier accounts stay at 30 RPM, 300 requests/day, 100,000 TPM and 1M TPD, which is the same evaluation sandbox as yesterday. Fair use and abuse review still exists, on request patterns not token counts. Anything unusual still gets a human looking at it.

For agents: x402

Every model is callable per-call with x402 (USDC on Base) at POST /v1/agent/chat/completions, with POST /v1/agent/quote to price a call in advance. Current quotas are in pricing.md and llms.txt.

Your ceiling is now what your card says. Build bigger.


JM
// The Founder

James Miano

CTO & ML Engineer at Lacesse Ventures. James specializes in low-overhead LLM quantization processes, custom ternary weights architectures, and localized server optimization.

Stop paying for overpriced round-trip latency

Why route queries over Western servers when you can use low-overhead hardware located in Nairobi? Save 87% on your monthly inference spend. No minimum credit limits. Pay with card or mobile money.

usagestatusdashboard