| Model | intelligence score | price for full suite | total tokens for suite |
| Sonnet 5.5_| 56 | $9,074 | 416M |
| Opus 5.5__| 58 | $8,708 | 260M |
| Sol 6.1____| 52 | $1,536 | _67M |
For a fixed set of tasks, the anthropic models (do score higher) but cost much more (both because per token prices are higher and because total token counts are higher).
So it isn't at all obvious that the anthropic subscription is more value. Anthropic's models cost more per token, which explains a large amount of the "value" difference you get from the plan. Then even if you look at tokens, you'er only getting ~3x more tokens from Anthropic which is not even enough to close the Opus-Sol6.1 efficiency gap.
This is a great first step for comparison, but I wish they paired this $ amount comparison to units of output.
Consider the data from https://artificialanalysis.ai :
| Model | intelligence score | price for full suite | total tokens for suite |
| Sonnet 5.5_| 56 | $9,074 | 416M |
| Opus 5.5__| 58 | $8,708 | 260M |
| Sol 6.1____| 52 | $1,536 | _67M |
For a fixed set of tasks, the anthropic models (do score higher) but cost much more (both because per token prices are higher and because total token counts are higher).
So it isn't at all obvious that the anthropic subscription is more value. Anthropic's models cost more per token, which explains a large amount of the "value" difference you get from the plan. Then even if you look at tokens, you'er only getting ~3x more tokens from Anthropic which is not even enough to close the Opus-Sol6.1 efficiency gap.