Every plan side by side — Go, GOAT, Pro, Max 10×/20×, Provider API & Teams. Credits that never expire, deals that apply automatically, zero markup.
Live from commandcode.ai/pricing — monthly credits, typical request volume, and effective usage with active deals.
On top of monthly credits, every subscription plan has two rolling windows — a 5-hour limit and a weekly limit — that smooth out bursts so a single session can't drain the whole month. Go, GOAT and Pro use 20% / 50% of their usage pool; Max and Team plans use 30% / 60%. Limits are measured in credit value, not request count.
| Plan | Price / mo | Monthly Credits | 5-Hour Limit | Weekly Limit | % of Monthly |
|---|---|---|---|---|---|
Rolling windows open on the first request after the previous one elapses and reset exactly one window-length later. Top-up and pay-as-you-go credits are never throttled — they spend first while you're over a window. Run /usage in the CLI for live 5-hour and weekly meters.
Applied automatically — no codes, no toggles. Stacks on top of your credit multiplier on every plan and even top-up credits.
Live from commandcode.ai/docs/resources/pricing-limits — 91 models, prices per 1M tokens. Deals marked ⚡, free models marked FREE.
| # | Model | Model ID | Provider | Context | Tier |
|---|---|---|---|---|---|
GOAT plan estimates (typical agent request: ~800 input tokens, ~50K cache reads, ~125–200 output tokens). Shows how far the $70 pool stretches per model.
Token prices per 1M tokens and monthly credit allowance per model on the GOAT and Pro plans.
| Model | Context | Input /1M | Output /1M | Cache Read /1M | Cache Write /1M | GOAT Credits | Req / 5h | Req / Week | Req / Month |
|---|
Rolling windows throttle only your included monthly credits. On-demand top-up credits are never throttled.
/extra — they roll over and never expireLaguna S 2.1, Ling, Glyph Cluster) keep working even with zero balance/usage in the CLI for live 5-hour & weekly metersSome providers restrict where their models can be used. Restrictions come from the provider, not Command Code — everything else keeps working.
GPT-5.6 Luna (OpenAI) — unavailable in ChinaGemini 3.7 Flash (Google) — unavailable in China/model in the CLI or StudioCommand Code negotiates capacity and runs 95–98% cache hit rates — savings come back to you as bigger credit allowances.