Pricing Engine v3
Pricing v3 updates relative model positioning across the catalog while keeping the same credit-based billing formula, model IDs, API compatibility, and existing API keys.
Overview
Amanai periodically reviews model rates as upstream capacity, model availability, and operating conditions evolve. Pricing v3 establishes a clearer relative ladder across model families and consistent component ratios throughout the token-billed catalog.
Component rates
Each model has an input multiplier, m_in. Generated output is billed at 5× the model input rate.
Cached input receives a 75% discount from the standard input rate. In other words, cached input is billed at 25% of m_in. The comparison table omits a separate cache column because the cache rate can be derived consistently for every model.
m_out = 5 × m_in and m_cache = 0.25 × m_in.
Pricing v2 vs Pricing v3
The table compares input and output multipliers side by side. Models are ordered by the Pricing v3 input rate, from highest to lowest. Values retain the precision used for billing.
| Model | Pricing v2 | Pricing v3 | ||
|---|---|---|---|---|
| m_in | m_out | m_in | m_out | |
 amanai/claude-fable-5 | 36.299373 | 181.496865 | 30 | 150 |
amanai/claude-opus-4.6 | 18.149686 | 90.74843 | 15 | 75 |
amanai/claude-opus-4.7 | 18.149686 | 90.74843 | 15 | 75 |
amanai/claude-opus-4.8 | 18.149686 | 90.74843 | 15 | 75 |
amanai/claude-opus-5 | 18.149686 | 90.74843 | 15 | 75 |
amanai/claude-sonnet-5 | 12.099791 | 60.498955 | 10.5 | 52.5 |
 amanai/gpt-5.5 | 4.3 | 21.5 | 10 | 50 |
amanai/gpt-5.6-sol | 6.34095 | 31.70475 | 10 | 50 |
amanai/claude-sonnet-4.6 | 12.099791 | 60.498955 | 9 | 45 |
 amanai/kimi-k3 | 15.08675 | 75.43375 | 9 | 45 |
amanai/gpt-5.6-terra | 5.97495 | 29.87475 | 7.5 | 37.5 |
 amanai/grok-4.5 | 5.73705 | 28.68525 | 6 | 30 |
 amanai/qwen3.8-max-preview | 6.03472 | 30.1736 | 6 | 30 |
amanai/gpt-5.6-luna | 6.15795 | 30.78975 | 5.625 | 28.125 |
 amanai/glm-5.1 | 4.2273 | 21.1365 | 5.6 | 28 |
amanai/glm-5.2 | 6.03472 | 30.1736 | 5.6 | 28 |
amanai/gpt-5.4 | 3.1 | 15.5 | 5 | 25 |
amanai/glm-5.0-turbo | 3.6234 | 18.117 | 4.8 | 24 |
amanai/glm-5v-turbo | 4.2273 | 21.1365 | 4.8 | 24 |
 amanai/hy3-preview | 3.32145 | 16.60725 | 4.5 | 22.5 |
amanai/qwen3.7-max | 4.228166 | 21.14083 | 4.425 | 22.125 |
amanai/glm-5.0 | 3.0195 | 15.0975 | 4 | 20 |
amanai/kimi-k2.6 | 4.34625 | 21.73125 | 3.8 | 19 |
amanai/kimi-k2.7 | 6.49755 | 32.48775 | 3.8 | 19 |
amanai/claude-haiku-4.5 | 5.499705 | 27.498525 | 3 | 15 |
amanai/kimi-k2.5 | 3.88875 | 19.44375 | 2.28 | 11.4 |
amanai/qwen3.7-plus | 3.500107 | 17.500535 | 2.24 | 11.2 |
 amanai/minimax-m3 | 2.67705 | 13.38525 | 2.1 | 10.5 |
 amanai/deepseek-v4-pro | 3.0195 | 15.0975 | 1.74 | 8.7 |
amanai/gpt-5.4-mini | 2.9 | 14.5 | 1.5 | 7.5 |
amanai/minimax-m2.7 | 1.8117 | 9.0585 | 1.2 | 6 |
amanai/deepseek-v4-flash | 1.5705 | 7.8525 | 0.87 | 4.35 |
amanai/deepseek-v4-flash-0731 | 1.5705 | 7.8525 | 0.87 | 4.35 |
amanai/qwen-lite | 0.003017 | 0.015085 | 0.003 | 0.015 |
Effective date
Pricing v3 takes effect at 00:00 UTC on August 10, 2026. Requests delivered before the effective time continue to use the applicable prior rate. Requests delivered from the effective time onward use Pricing v3.
Before the update. Pricing v2 remains in effect until 00:00 UTC on August 10, 2026, giving you time to review the updated rates and adjust your usage preferences before Pricing v3 begins.
Billing impact
Existing credit balances do not need to be converted or migrated. Historical usage and previously settled charges are not recalculated. Actual credit consumption continues to depend on model choice, input and output token counts, cached-input usage, and the minimum charge for billable requests.
After the effective date, Billing & pricing and the credit calculator show the current rates used for new usage.