Usage

Last updated

Since September 2025: 33.13B tokens across 249 active days, mostly GPT-5.6 Sol via Claude Code. Current streak: 166 days.

API equivalent
US$27.8K
Tokens
33.13B
Active days
249
Models
56

The API equivalent of my AI coding agents at provider list prices. Tokens, cost, and reasoning effort. Not what I paid. Aggregates only.

Model character

Cache reuse, blended price, how often each model runs, and when it last ran. Open a model's profile for its full history.

ModelCache hit$ / 1M tokensActive daysLast used
GPT-5.6 Sol
97%
US$0.732824/09/2026
Claude Opus 5
97%
US$0.795222/09/2026
Claude Opus 4.8
95%
US$0.926001/08/2026
Claude Opus 5.5
97%
US$0.40729/09/2026
Claude Fable 5
96%
US$1.604923/08/2026
GPT-5.5
93%
US$0.916801/08/2026
GPT 5.5 (Fast)
94%
US$2.083901/07/2026
Claude Opus 4.7
92%
US$1.149529/09/2026

Where the money goes

Blended price per million tokens against volume, per model. Bigger dots mean more messages. Both axes are log scale.

Biggest spenders: GPT-5.6 Sol, US$4,477.95 at US$0.73 per million tokens; Claude Opus 5, US$4,201.75 at US$0.79 per million tokens; Claude Opus 4.8, US$4,086.80 at US$0.92 per million tokens; Claude Fable 5, US$3,835.83 at US$1.60 per million tokens; GPT 5.5 (Fast), US$3,824.24 at US$2.08 per million tokens. Every model's figures are in the Explorer table below.

Cache & effort

Prompt tokens served from cache, what that saved against the full input price, and how hard I pushed the models.

Cache hit rate
95%
Saved by caching
US$153.7K
Sep 2025Sep 2026

Effort levels

2,398 of 2,855 sessions classified.

  • Minimal

    13 (1%)

  • Low

    201 (8%)

  • Medium

    722 (30%)

  • High

    303 (13%)

  • Extra high

    972 (41%)

  • Max

    31 (1%)

  • Ultra

    152 (6%)

  • Mixed

    4 (0%)

Token mix

All 33.13B tokens by category

  • Cache read

    31.43B (95%)

  • Cache write

    793.77M (2%)

  • Input

    775.73M (2%)

  • Output

    103.38M (0%)

  • Reasoning

    28.58M (0%)