Cheap open models do the grunt work. You pay a flat price for a weekly limit, or pay API prices for what you use. No "unlimited", ever.
No plan. Prepaid balance, every model.
Minimum top-up $5
For the daily grunt work.
Billed monthly
For the agent that never sleeps.
Billed monthly
Every plan has a hard cap. Limits are estimates, not promises. A "heavy agent session" is 2 to 3 hours of agentic coding. What is that?
The meter is a cost budget, not a token count. One heavy session fills this much of a Servant week (illustration, estimate).
USD per 1M tokens. Used by Pay as you go and for going over a plan limit. Launch prices, may change.
| Alias | Model | Input | Cached | Output | Context | Heavy session (est.) | Badge |
|---|---|---|---|---|---|---|---|
| Normal models | |||||||
tco/deepseek-flash | DeepSeek V4.1 Flash | $0.05 | $0.007 | $0.55 | 1M | ~$0.28 | - |
tco/qwen-flash | Qwen 3.8 Flash | $0.17 | $0.018 | $0.52 | 1M | ~$0.59 | - |
tco/minimax-flash | MiniMax M3.1 Flash | $0.17 | $0.017 | $0.66 | 1M | ~$0.61 | - |
tco/glm | GLM 5.3 | $0.09 | $0.084 | $3.96 | 1M | ~$2.06 | - |
tco/kimi-code | Kimi K2.7 Code | $0.66 | $0.165 | $3.85 | 262K | ~$4.00 | - |
| Refusal-free models | |||||||
tco/deepseek-flash-rf | DeepSeek V4.1 Flash, refusal-free | $0.13 | $0.028 | $0.50 | 1M | ~$0.67 | refusal-free |
tco/glm-flash-rf | GLM 5.3 Flash, refusal-free | $0.13 | $0.033 | $0.55 | 1M | ~$0.75 | refusal-free |
tco/glm-rf | GLM 5.3, refusal-free | $1.10 | $0.165 | $2.20 | 1M | ~$4.30 | refusal-free |
2 to 3 hours of agentic coding: about 15M input tokens (90% of them cached) and 0.2M output tokens.
It is an estimate. Your real use may be lighter or heavier.
Two choices. Keep going at API prices from a prepaid balance (you opt in first), or wait for the weekly reset.
We never bill you past a cap you did not set.
It answers instead of refusing. It is the same model with the refusal direction removed (this is called "abliteration").
It is no smarter than the base model. Same skills, same mistakes.
It does not decide what is legal. You do, and our usage policy still applies.
Legal use only: security work, red-teaming your own AI, research, fiction. Read the usage policy.
On US hosts. We never route to China-based APIs.
Yes, any time. You keep your plan until the end of the period you paid for.
Because it is never true. Every plan has a weekly limit and a hard cap, and we show the limit to you before you pay.