Coding Plan Week
Per 7 days
≈ $1.67
(charged Rp 30.000)
API balance worth $3.33
- Balance active for 7 days
- Access to 7 coding models
- No request limit
- No short-window cap
- ~6,363 coding requests per 7 days
Coding Plan
One plan. Every coding model.
Pay once, get an API balance worth more than the price. The balance works on every eligible model, from flash to frontier, and isn't split by vendor. When the window ends, just buy again.
Coding Plan Month
Per 30 days
≈ $4.94
(charged Rp 89.000)
API balance worth $13.89
- Balance active for 30 days
- Access to 9 coding models
- No request limit
- Max $2.78 per 5 hours, $6.94 per 7 days
- ~11,842 coding requests per 30 days
Coding Plan Pro
Per 30 days
≈ $8.33
(charged Rp 150.000)
API balance worth $27.78
- Balance active for 30 days
- Access to 15 coding models
- No request limit
- Max $5.56 per 5 hours, $13.89 per 7 days
- ~6,307 coding requests per 30 days
Usage calculator
Pick a plan and a model, then adjust the shape of your requests. The math uses the same per-model prices as billing.
On Coding Plan Week, you pay ≈ $1.67 (charged Rp 30.000) and get an API balance worth $3.33 for 7 days, one balance for all 7 models in this plan.
Plan
7 models in this plan
800
200
50K
Total requests / 7 days
10K
on Qwen 3.8 Flash: $3.33 ÷ $0.0003 per request
Tokens / 7 days±50%, cache varies
Total510M
Cache500M
Input8M
Output2M
Plan mix average: 6,363 requestsif requests are split evenly across 7 models in Week
Cost / request$0.0003$0.0003 rounded up per request
Price / 1M tokens
Input$0.05
Output$0.16
Cache$0.005
No burst limit: you can spend the whole balance at any time within 7 days.
- Quick fix2,000single-pass edit
- Bug fix400patch + verification
- Small CLI133built from scratch
- Feature PR67multi-file change
- Side project25weekend project
What you can build per 7 days, a rough estimate on Qwen 3.8 Flash with 10K requests.
Buy WeekAn estimate, not a promise. Real usage depends on the model, prompt length, and how long a session runs: every new turn re-reads the whole context, so long sessions use up the balance faster. Exact per-request figures are on the Usage page of the dashboard.
What makes it different.
- One wallet, every model. Not split by vendor: leftover flash balance can be used on frontier models, and the other way around.
- No request limit. The only limits are your balance across 7 eligible models and a rate limit of 60 req/min per key.
- Cache hits are discounted. Cached prompts are billed at 10-25% of the input price, so agentic requests cost far less than a rough estimate suggests.
- No silent balance loss. When the window ends, the plan stops. Your pay-as-you-go balance stays intact and is never touched by the plan.
Request estimate = the average across all models in that plan, for a typical coding request: 800 new input tokens, 200 output tokens, 50,000 cache read tokens. Long sessions with big contexts produce smaller numbers; adjust it yourself in the calculator and check real usage in the dashboard. Unused plan balance can't be cashed out and doesn't carry over between windows.