On this page⌄

Docs · Plans

Starter plan

A dollar to try every model.

Price
$1 / month
Credits
$5 / month
Requests
~340
Models
All 21

Requests included#

The Starter plan includes $5 of credits a month, spent at each model maker's official API price. On Claude Sonnet 5.5, that is about 340 requests.

A request is one agent step: the conversation so far (about 50K tokens, read from the prompt cache), about 1K new tokens and a 220-token answer. On Claude Sonnet 5.5 that step costs $0.0147, so $5 ÷ $0.0147 ≈ 340. Cheaper models stretch the same credits further:

DeepSeek V4.1 Flash
5.7K
GLM 5.3 Flash
2.8K
DeepSeek V4 Pro
1.1K
Claude Haiku 4.5
680
GPT-6.1 Sol
543
Claude Sonnet 5.5
340
GLM 5.2
325
Claude Opus 5.5
257
GPT-5.6 Sol
176
Requests per month on the Starter plan, by model

Short questions use fewer credits; big files and long reasoning use more. The CCode sidebar shows what each session has spent as you go.

Models#

Credits work on all 21 models from Anthropic, OpenAI, DeepSeek and Z.ai. You can spend the whole $5 on one model or mix them. The four below are the ones most people start with; the full price table lists the rest.

Claude Opus 5.5

Anthropic

Hard problems: large refactors, subtle bugs, architecture.

Input
$4
Cached
$0.20
Output
$20

$5 covers ~257 requests

GPT-6.1 Sol

OpenAI

Everyday coding, reviews and tests, at about half Opus's cost per step.

Input
$2
Cached
$0.10
Output
$10

$5 covers ~543 requests

DeepSeek V4.1 Flash

DeepSeek

Fast, cheap iterations and long agent loops.

Input
$0.30
Cached
$0.006
Output
$1.20

$5 covers ~5.7K requests

GLM 5.2

Z.ai

A strong open model for routine work at a low price.

Input
$1.40
Cached
$0.26
Output
$4.40

$5 covers ~325 requests

Prices in USD per 1M tokens.

How usage refreshes#

Your usage refills on three schedules: every 5 hours, every week and every month. That spreads your $5 across the whole month, so a long session today still leaves plenty for tomorrow.

Every 5 hours
$1
Every week
$2.50
Every month
$5
Usage available in each window

Up to 8 requests can run at once, so helpers can work in parallel.

Prompt caching#

Long sessions stay affordable: whatever the model has already seen in your conversation is billed at its cached rate, up to 95% below the input price, so even deep into a task each step costs just cents.

Billing#

Your plan renews every month and your credits refresh with it. Need more? Pro gives you $70 of credits for $10 a month.

Starter · $1/month for $5 of credits

Get Starter