Calculator

How many lanes do you need?

Every lane is the same, and bigger models take more of them. Pick a plan, then drag how many requests of each model you run at once. If they don’t fit, you’ll see how many lanes you’re short and what fixes it.

Pre-launch. Accounts are open; the API and checkout open soon.

Plan
Kimi K36 lanes each
1 at once6 lanes
GLM 5.32 lanes each
2 at once4 lanes
MiMo V2.51 lane each
1 at once1 lane
GLM 5.3 Flash1 lane each
2 at once2 lanes
DeepSeek V4.1 Flash1 lane each
1 at once1 lane
gpt-oss-120b1 lane each
0 at once0 lanes

Turbo is 2x the lanes for 2–5x the throughput, depending on model: a request takes Turbo lanes, or twice its lanes when no Turbo lanes are free. An estimate for planning: choosing Turbo per request arrives with the API.

Pro · 16 lanes14 busy · 2 free
  • free lane
  • Turbo lane
  • running at Turbo
  • missing lane

Fits Pro, with 2 lanes to spare.

To run this, per month

How it counts

One sum, no meter.

A request holds its lanes until its last token, then they’re free for the next. What doesn’t fit waits its turn; nothing is billed extra. Need more room than a plan has? Add lanes from $6 a month.

Compare plans
Modellanes per request
MiMo V2.51 lane
GLM 5.3 Flash1 lane
DeepSeek V4.1 Flash1 lane
gpt-oss-120b1 lane
GLM 5.32 lanes
Kimi K36 lanes
At Turbox2

About 150–400 tok/s instead of up to 150 tok/s. On Turbo lanes, or twice as many lanes combined.