Simple plans, hard limits.
Try it with no account, then keep going free every day. Every plan has every feature: undo, approvals, derby, swarm, brain and voice. Plans differ in how much you can use and which models, and you're never charged past your limit.
Free
Try it properly on real work.
$0
- Try it first with no account
- A daily allowance that resets every 24 hours
- Models up to $5 per million output tokens
Builder
RecommendedFor daily use on real projects.
$15/ month
- About 7× the Free allowance, reset weekly
- Models up to $20 per million output tokens
- Includes Claude Opus 5.5
Power
For heavy use, swarms and derbies.
$65/ month
- About 4× Builder, reset weekly
- Every model, including the priciest reasoning ones
- Priority support
How usage works
Each plan has an allowance of model use: Free per day, Builder and Power per week. It's measured by what the work actually costs, so fast models stretch it much further than the biggest ones. Roughly how many typical tasks (a fix or small feature, about 15 model calls) fit:
| Model | Free, per day | Builder, per week | Power, per week |
|---|---|---|---|
| Claude Haiku 5.5 | about 6 | about 345 | about 1,521 |
| Qwen3 Coder 480B A35B | about 1 | about 97 | about 430 |
| Gemini 3.8 Flash | less than 1 | about 34 | about 153 |
| Claude Sonnet 5.5 | Not included | about 17 | about 76 |
| Claude Opus 5.5 | Not included | about 8 | about 38 |
Estimates from real sessions at today's prices; long tasks in big codebases use more. Darce and the dashboard show how much of your allowance you've used and when it resets.
Which models each plan includes
Plans include models by their price per million output tokens: Free up to $5, Builder up to $20, Power everything. Prices are OpenRouter's, refreshed daily.
Free
| Model | Output price |
|---|---|
| Claude Haiku 5.5 | $0.50 / M tokens |
| GPT-6 Luna | $0.50 / M tokens |
| gpt-oss-120b | $0.17 / M tokens |
| Gemini 3.8 Flash | $3.75 / M tokens |
| Gemini 3.5 Flash Lite | $2.50 / M tokens |
| Gemma 4 31B | $0.34 / M tokens |
| Qwen3 Coder 480B A35B | $1.00 / M tokens |
| Qwen3 Coder Next | $0.80 / M tokens |
| Qwen3 Coder Plus | $3.25 / M tokens |
| Qwen3 Coder Flash | $0.97 / M tokens |
| Qwen3 Coder 30B A3B Instruct | $0.28 / M tokens |
| Qwen3.8 Flash | $0.47 / M tokens |
| DeepSeek V4 Pro 0423 | $1.90 / M tokens |
| DeepSeek V4.1 Flash | $1.20 / M tokens |
| DeepSeek V3.2 | $0.80 / M tokens |
| Kimi K2.7 Code | $3.35 / M tokens |
| Kimi K2.6 | $2.45 / M tokens |
| Kimi K2 Thinking | $2.50 / M tokens |
| GLM 5.3 | $4.80 / M tokens |
| GLM 5.3 Flash | $0.50 / M tokens |
| GLM 5.1 | $3.04 / M tokens |
| Grok Build 0.1 | $2.00 / M tokens |
| MiniMax M3 | $1.20 / M tokens |
| MiniMax M2.7 | $0.84 / M tokens |
| Devstral 2 2512 | $2.00 / M tokens |
| Codestral 2508 | $0.90 / M tokens |
| Mistral Large 4 | $2.09 / M tokens |
Builder adds
| Model | Output price |
|---|---|
| Claude Opus 5.5 | $20.00 / M tokens |
| Claude Sonnet 5.5 | $10.00 / M tokens |
| Claude Sonnet 4.5 | $15.00 / M tokens |
| GPT-6.1 Sol | $10.00 / M tokens |
| GPT-6 Sol | $10.00 / M tokens |
| GPT-5.6 Sol | $10.00 / M tokens |
| GPT-5.3-Codex | $14.00 / M tokens |
| Gemini 3.1 Pro Preview | $12.00 / M tokens |
| Qwen3.8 Max (0902) | $6.00 / M tokens |
| Kimi K3 | $13.50 / M tokens |
| Grok 4.7 | $6.00 / M tokens |
Power adds
| Model | Output price |
|---|---|
| Claude Fable 5.1 | $50.00 / M tokens |
| Claude Opus 5 | $25.00 / M tokens |
| GPT-6 Astra | $50.00 / M tokens |
| GPT-5.5 | $30.00 / M tokens |
Questions
How is usage measured?
By what Darce's model calls actually cost, at the model's own price. A task on a fast model like Claude Haiku or Qwen3 Coder uses a small slice of your allowance; the same task on Claude Opus uses much more. Darce shows how much of your allowance you've used, never a bill.
When does my allowance reset?
Free resets every 24 hours from when you started using it. Builder and Power reset every 7 days. Darce and the dashboard show when.
What happens when I run out?
Darce stops and tells you when it resets. You are never charged extra and there are no overages. Switching to a cheaper model with /model makes what's left go further.
Do /derby and /swarm use more?
They run several agents, so they use several times what one run would: a derby of three models uses about three times as much.
Can I cancel any time?
Yes. Manage or cancel from the dashboard's Billing page; you keep the plan until the end of the period you paid for.
Is there a free trial of the paid plans?
Darce is in early beta. If you got a beta code, run darce redeem <code> to get Power free for 60 days. When it ends you go back to Free; nothing is charged.