Pricing

One subscription, every model

Plans include monthly token capacity that works across every model in your tier. Payments powered by Waffo Pancake.

Lite

Try agents without commitment

$4.90/month

Models

qwen3.8-flashglm-5.3-flash

~24–28M tokens/mo

estimated usage — agent-style mix*

2 flash models
Credit resets monthly
Request logs & analytics
Community support

Basic

Casual daily use

$9.90/month

Models

qwen3.8-flashglm-5.3-flashdeepseek-v4.1-flash

~23–55M tokens/mo

estimated usage — agent-style mix*

3 flash models
Credit resets monthly
Request logs & analytics
Community support

Starter

Regular agent workloads

$17.86/month

Models

qwen3.8-flashglm-5.3-flashdeepseek-v4.1-flash

~45–111M tokens/mo

estimated usage — agent-style mix*

3 flash models
Credit resets monthly
Request logs & analytics
Community support

Pro

Popular

Heavy agents & production

$53.57/month

Models

All flash modelsMiniMax-M3glm-5.3qwen3.8-max

~29–324M tokens/mo

estimated usage — agent-style mix*

6 models, incl. mid-tier
Credit resets monthly
Request logs & analytics
Priority support

Max

Max capacity, strongest models

$142.86/month

Models

Every model on the gatewaykimi-k3

~28–832M tokens/mo

estimated usage — agent-style mix*

All 7 models, incl. kimi-k3
Credit resets monthly
Request logs & analytics
SLA & onboarding support

How token billing works

Every request consumes credit from your plan based on three token types. Cache hits are dramatically cheaper — agent-style workloads that reuse context get the most out of every plan.

Input tokens

Base rate

Everything you send to the model — instructions, code, documents, chat history. Billed at the model's base rate.

Cache reads

2.5–29% of input price

Repeated context (like an agent's system prompt) hits the cache and is billed at a steep discount. Agents send 60–90% of input as cache — that's why they're cheap to run.

Output tokens

3.4–5× input price

What the model generates — answers, reasoning, code. The most expensive token type, so long outputs burn credit fastest.

Monthly capacity per plan

Millions of tokens per month. Ranges depend on which model in your tier you use — flash models go furthest.

Plan	Input only	Cache only	Output only	Typical usage*
Lite	20M	70–161M	±6M	24–28M
Basic	16–40M	140–804M	4–12M	23–55M
Starter	32–80M	280M–1.6B	8–24M	45–111M
Pro	16–235M	96M–4.7B	5–70M	29–324M
Max	24–603M	244M–12.1B	5–179M	28–832M

Input/cache/output-only columns are theoretical extremes — real usage is always a mix. *Typical usage assumes an agent-style mix: every 1M input tokens comes with ~0.7M cache reads and ~0.1M output tokens. Capacity resets monthly while your subscription is active.