Which Agent
Free tool

Cost & allowance calculator

Agents use 5–30× the tokens of a chatbot reply per task ✓ verified. Set your workload and see which plans could cover it. Prices as of 20 Oct 2026.

Estimated tokens per week
—

Only Meta publishes a token allowance (“Muse tokens”, its own metering unit). Everyone else meters weekly or by credits without publishing numbers. Real agent runs that browse and retry can use far more than this simple model.

Muse: your estimate vs the published weekly allowance

Every plan: price per task

AgentPlanPrice / monthPublished allowance≈ price per task at your volumeStatus
MuseFree$0100M tokens/week—◐
MusePower$20500M tokens/week—✓
MuseMaximum$1003B tokens/week—✓
DotsPro 100$100not published—✓
DotsPro 200$200not published—✓
DotsBusiness Premium$125not published—◐
Grok BotPremium$30not published—✓
Grok BotSuper$100not published—✓
Grok BotUltra$200not published—✓
Gemini SparkAI Pro$19.99not published—✓
Gemini SparkAI Ultra 5×$99.99not published—✓
Gemini SparkAI Ultra 20×$199.99not published—✓
Claude (Cowork / Code)Pro$20not published—✓
Claude (Cowork / Code)Max 5×$100not published—✓
Claude (Cowork / Code)Max 20×$200not published—✓
Microsoft AutopilotCopilot Creditsusage-basednot published—◐
Siri AIIncluded$0not published—✓
Perplexity ComputerPro$20not published—◐
Perplexity ComputerMax$200not published—✓
InstinctInvite$0not published—✓
Cue (Manus)Early access$0not published—◐

Price per task = monthly price ÷ (weekly tasks × 4.33). It ignores allowances, so it understates the cost of heavy use and overstates it for light use. Microsoft Copilot Credits list at $0.01 each ◐, but credits per Autopilot task aren’t published.