Pro+ vs. usage base pricing

[Artemonim]

Extreme token usage

The count here includes cached read tokens, which get included based on all prior input tokens at every tool call. Cached input tokens are 10x cheaper than input tokens. If you hover over the token count, it should give you a breakdown by token type.

@Artemonim

Model	Est. Input CPM ($/M tokens)	Est. Output CPM ($/M tokens)
Grok-4	~$3.11	~$15.57
Gemini 2.5 Pro	~$3.98	~$19.90
Claude 4 (Sonnet)	~$4.30	~$21.50
Claude 4 Thinking	~$7.49	~$37.45
o3-pro	~$28.85	~$86.54
o4-mini	~$0.17	~$0.68
auto	~$3.09	~$6.18

Recalculated API Cost Per Million Tokens

The new estimates for cost per million tokens (CPM), factoring in the cost of Cache Read, are as follows:

Model	Est. Input CPM ($/M tokens)	Est. Output CPM ($/M tokens)
Grok-4	~$3.21	~$16.05
Gemini 2.5 Pro	~$2.26	~$6.78
Claude 4 (Sonnet)	~$4.89	~$24.45
Claude 4 Thinking	~$3.92	~$19.60
o3-pro	~$25.35	~$76.05
o4-mini	~$1.49	~$5.96
auto	~$2.54	~$5.08

• Note on Calculation: These estimates are derived by treating Cache Read as a billable event costing 10% of the standard input price. A common industry price ratio between output and input tokens (e.g., 5:1 for Claude, 3:1 for Gemini) was assumed to solve for the individual rates. The models with a high volume of Cache Read tokens (like Gemini and Claude) see a significant shift in their effective cost structure compared to the previous calculation.