Every AI coding subscription measured against the vendor's own API price. These numbers are regenerated from the catalogue on every deploy, because prices and caps move constantly and a report written once is wrong within weeks.
18 of 18 plans have a ceiling somebody can state. The rest publish a price and a multiplier and nothing in between.
| Plan | Price | Ceiling | Multiple | Basis |
|---|---|---|---|---|
| Moonshot AI Moderato | $19/mo | $1,850/mo | 97.4x | estimate |
| Moonshot AI Allegretto | $39/mo | $3,699/mo | 94.8x | estimate |
| Z.ai Coding Plan Max | $168/mo | $15,484/mo | 92.2x | estimate |
| Z.ai Coding Plan Pro | $80/mo | $6,636/mo | 83.0x | estimate |
| Moonshot AI Allegro | $99/mo | $7,398/mo | 74.7x | estimate |
| Z.ai Coding Plan Lite | $18/mo | $1,106/mo | 61.4x | estimate |
| MiniMax Token Plan Ultra | $120/mo | $6,562/mo | 54.7x | estimate |
| MiniMax Token Plan Max | $50/mo | $2,677/mo | 53.5x | estimate |
| xAI SuperGrok | $30/mo | $1,454/mo | 48.5x | measured |
| MiniMax Token Plan Plus | $20/mo | $892/mo | 44.6x | estimate |
| OpenAI ChatGPT Go | $8/mo | $199/mo | 24.9x | estimate |
| OpenAI ChatGPT Plus | $20/mo | $497/mo | 24.9x | measured |
| OpenAI ChatGPT Pro | $100/mo | $2,387/mo | 23.9x | measured |
| xAI X Premium+ | $40/mo | $872/mo | 21.8x | estimate |
| xAI SuperGrok Heavy | $300/mo | $5,816/mo | 19.4x | estimate |
| Anthropic Max 20x | $200/mo | $2,679/mo | 13.4x | from published cap |
| Anthropic Max 5x | $100/mo | $1,339/mo | 13.4x | from published cap |
| Anthropic Pro | $20/mo | $161/mo | 8.1x | from published cap |
Blended list API rate, weighting input three to one against output because a coding agent reads far more than it writes. The spread between cheapest and dearest is 21.4x.
| Model | Vendor | In | Out | Blended |
|---|---|---|---|---|
| MiniMax M3 | MiniMax | $0.30 | $1.20 | $0.53 |
| Kimi K2.7 Code | Moonshot AI | $0.95 | $4.00 | $1.71 |
| GLM-5.3 | Z.ai | $1.40 | $4.40 | $2.15 |
| Grok 4.6 | xAI | $2.00 | $6.00 | $3.00 |
| GPT-5.6 Terra | OpenAI | $2.00 | $12.00 | $4.50 |
| Claude Sonnet 5 | Anthropic | $3.00 | $15.00 | $6.00 |
| Kimi K3 | Moonshot AI | $3.00 | $15.00 | $6.00 |
| Claude Opus 5 | Anthropic | $5.00 | $25.00 | $10.00 |
| GPT-5.6 Sol | OpenAI | $5.00 | $30.00 | $11.25 |
0 plans publish no cap anybody can convert into a number: . These are blank on the site rather than estimated, because a plausible guess presented as a measurement is worse than an admitted gap.
Each one needs somebody on that plan to run the measurement skill. It reads token counts the agent already writes to disk and takes about a minute.
from published cap are exact: the vendor publishes the allowance and the rate it meters at, so the ceiling is arithmetic rather than inference.measured come from a real window, either one the user exhausted or one the agent reported its own quota usage for.estimate are a published quota times a stated assumption about what a unit of it costs in tokens. The assumption is printed with the figure so it can be argued with.scaled are a p90 of measured windows scaled by a published multiplier, and are the weakest of the three.