Capacity
What $5 buysdepends on the class.
Each class is a published promise: which models it serves, at what reference price, since when.
$5 gets you
≈ 7.83M tokens
$0.64 per million tokens · reference price
Blended price $0.24 to $0.53 per million tokens
The blended price counts three input tokens for each output token.
+ 40% rework margin for this tier. Weaker tiers need more retries and fixes, so their margin is higher. A declared default for each tier, not a measurement yet.
≈ 100K tokens per model hour · published estimate declared Oct 7 (sample so far: 0 of 50 active hours, 0 of 5 keys)
Models in this class · 5
- @cf/google/gemma-4-26b-a4b-it
- @cf/meta/llama-4-scout-17b-16e-instruct
- @cf/openai/gpt-oss-120b
- @cf/qwen/qwen3-30b-a3b-fp8
- @cf/zai-org/glm-5.3-flash
$5 gets you
≈ 991.4k tokens
$5.04 per million tokens · reference price
Includes Essential models
Blended price $1.88 to $4.20 per million tokens
The blended price counts three input tokens for each output token.
+ 25% rework margin for this tier. Weaker tiers need more retries and fixes, so their margin is higher. A declared default for each tier, not a measurement yet.
≈ 100K tokens per model hour · published estimate declared Oct 7 (sample so far: 0 of 50 active hours, 0 of 5 keys)
Models in this class · 4
- @cf/deepseek-ai/deepseek-v4-pro-0813
- @cf/moonshotai/kimi-k2.6
- @cf/moonshotai/kimi-k2.7-code
- @cf/zai-org/glm-5.3
Includes Essential and Advanced models
Reference is not what you are charged.
Estimates only. You are charged for what is actually used.
Legacy classes
They keep serving capacity already sold and take no new cup. Each page stays up: whoever funded a cup on that class must keep finding the promise that applied at the time.
Serves capacity already sold. Takes no new cup.
Backers fund compute. What comes back is capacity for the cup, never money.