Skip to main content

​
Plans & Usage

Every plan includes a request allowance per usage window (for example, 30 requests per 5 hours on Promo, 800 on Pro), a weekly request cap that resets Monday 00:00 UTC, and a per-minute burst cap (10 per minute on Promo, up to 25 on paid tiers). All paid plans include coding tools and autonomous agents through one subscription. When you need more, on-demand overage kicks in — no hard cutoff. See live plan details and pricing on the Pricing page.

​
Plan limits at a glance

All limits are per account — your API keys share one window allowance, weekly cap, and burst budget. Allowances below are counted in requests’ worth — most chat models cost 1.0 per call (see Model Request Ratios). Window allowances reset in full on a fixed schedule — your exact reset time is shown in the dashboard under Billing. Weekly caps reset every Monday 00:00 UTC. Existing subscribers on legacy pricing keep their current rate until they choose to switch. On-demand overage and credits can extend your window allowance and weekly cap; the per-minute burst limit is a hard cap and is never extended by credits. Prices are shown in USD — local pricing, annual billing, and any current promotions are shown live on the Pricing page, which is authoritative if it differs from this table.

​
What’s included

Each plan includes:
  • A request allowance per usage window (shown on plan cards as requests per 5 hours)
  • A weekly request cap (resets Monday 00:00 UTC)
  • Access to the curated model catalog, including the latest open-source models (a small number of premium models may be limited to specific plans — see the Pricing page)
  • On-demand overage at catalog model rates
  • Optional credit top-ups for predictable costs
  • Paid plans: autonomous agents and supported coding tools on one subscription
Higher tiers include more requests per window, a larger weekly cap, and additional features such as usage analytics and priority support (Pro, Pro+, Max, and Max+).

​
Promo plan ($1/mo)

The Promo plan ($1/month) gives developers an entry point for testing and lightweight daily workloads:
  • 2-day full catalog trial: access all standard catalog models during the initial 48 hours.
  • Post-trial included models: ongoing access to Qwen3.8 27b and Llama 3.2 3B at a discounted 0.1 request cost (10 quota units) per call.
  • How the math works:
    • Per 5-hour window: 30 allowance units ÷ 0.1 ratio = 300 requests every 5 hours.
    • Per week: 420 allowance units ÷ 0.1 ratio = 4,200 requests per week.
  • Edge routing: Served via our distributed edge infrastructure for dependable throughput.

​
Refunds

All subscription and credit purchases are final and non-refundable, except where a refund is required by applicable law. We recommend testing the service on the Promo plan before purchasing a paid subscription. See the Terms of Service for the full refund policy and the 30-day billing-dispute window.

​
What counts as a request?

A request counts toward your allowance when the final status recorded is:
  • success
  • client_error (4xx from client or bad request)
Upstream provider errors and capacity unavailable situations do not count against your quota.

​
Model multipliers

How much of your allowance a call uses depends on which model you call. Auto and flash models are the 1.0 floor — one request’s worth each. GLM-5.2 costs 2.0. Mid-tier models sit between 1.0 and 2.0; premium models cost more. So a plan with 800 requests per window gives you 800 Auto calls, 533 Nemotron-3-120B calls, 500 Kimi K2.7 Code or Kimi K2.6 calls, 470 DeepSeek-V4-Pro-0813 calls, or 400 GLM-5.2 calls. Models including GLM-4.7-Flash, Nemotron-3-120B, Kimi K2.6, and Qwen3.8 27b are served via our distributed edge infrastructure for reliable throughput. Model Request Ratios has the per-plan request counts for every model. Every response carries the cost of that call in the X-Quota-Cost header, and GET /v1/models reports each model’s quota_multiplier, so an agent can choose at runtime without hardcoding this table.

​
Long-context surcharge

A GLM-5.3 request whose prompt exceeds 262,144 tokens costs 4 requests instead of 2. It is the only chat model with a context surcharge, and it is measured on upstream-reported input tokens after the call completes. A long reply to a short prompt still costs 2. The X-Quota-Cost response header uses a pre-flight estimate and may read lower until settlement; your dashboard quota is exact. If you routinely work above 262k tokens, DeepSeek-V4-Pro-0813 (also accepted as DeepSeek-V4-Pro) and MiniMax M3 both have larger context windows and carry no surcharge.

​
Going over your limit

When you exceed your included requests in the current window or your weekly cap, you can continue using the API:
  1. On-demand usage — Overage is billed at catalog model rates.
  2. Credit top-ups — Add a dollar balance in the dashboard for predictable spending.
  3. Upgrade — Move to a higher tier for a larger allowance.
There is no hard cutoff when credits or on-demand usage are enabled. Your usage window resets on a fixed schedule — your full allowance comes back at each reset, and your exact reset time is shown in the dashboard under Billing. The weekly cap resets every Monday 00:00 UTC.

​
Burst protection

To keep the platform responsive for everyone, Openference also applies a per-minute burst limit on each account (all of your API keys share one burst budget). This is separate from your per-window request allowance, it is a hard cap (requests beyond it return 429), and it is not bypassed by credits — only your per-window allowance can be extended with on-demand credits. For typical interactive IDE use, you will stay well within the burst limit. Paid tiers offer higher burst caps (15–25 requests per minute depending on plan). If you are hitting the burst limit regularly, upgrade your plan or add credits for on-demand overage. See Rate Limits & Fair Usage for technical details on burst limits and 429 responses.

​
Track your usage

View your current usage, remaining allowance, and billing history in the dashboard under Billing.