feat: tiered pricing based on prompt-side tokens #3157
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,92 @@ | ||
| package ratio_setting | ||
|
|
||
| import ( | ||
| "github.com/QuantumNous/new-api/types" | ||
| ) | ||
|
|
||
| // TieredPricingTier defines a pricing tier for a model. | ||
| // When prompt-side tokens >= Threshold, the ratios in this tier override the base ratios. | ||
| // Tiers should be sorted by Threshold in ascending order. | ||
| type TieredPricingTier struct { | ||
| Threshold int `json:"threshold"` | ||
| ModelRatio float64 `json:"model_ratio"` | ||
| CompletionRatio float64 `json:"completion_ratio"` | ||
| CacheRatio float64 `json:"cache_ratio"` | ||
| } | ||
|
|
||
| // Default tiered pricing for models that have context-based pricing tiers. | ||
| // Key is model name, value is list of tiers (sorted by threshold ascending). | ||
| // The first tier (threshold=0) is not needed here — it uses the base ratios from modelRatioMap etc. | ||
| // Only tiers above the base need to be listed. | ||
| var defaultTieredPricing = map[string][]TieredPricingTier{ | ||
| // gpt-4.1: base $2/1M input, $8/1M output; 200K+: $4/1M input, $8/1M output (doubled input, same output) | ||
| "gpt-4.1": {{Threshold: 200000, ModelRatio: 2.0, CompletionRatio: 2.0, CacheRatio: 0.5}}, | ||
| "gpt-4.1-2025-04-14": {{Threshold: 200000, ModelRatio: 2.0, CompletionRatio: 2.0, CacheRatio: 0.5}}, | ||
| "gpt-4.1-mini": {{Threshold: 200000, ModelRatio: 0.4, CompletionRatio: 4.0, CacheRatio: 0.5}}, | ||
| "gpt-4.1-mini-2025-04-14": {{Threshold: 200000, ModelRatio: 0.4, CompletionRatio: 4.0, CacheRatio: 0.5}}, | ||
| "gpt-4.1-nano": {{Threshold: 200000, ModelRatio: 0.1, CompletionRatio: 4.0, CacheRatio: 0.5}}, | ||
| "gpt-4.1-nano-2025-04-14": {{Threshold: 200000, ModelRatio: 0.1, CompletionRatio: 4.0, CacheRatio: 0.5}}, | ||
| // gpt-5: base $1.25/1M input, $10/1M output; 272K+: $2.5/1M input, $15/1M output | ||
| "gpt-5": {{Threshold: 272000, ModelRatio: 1.25, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| "gpt-5-2025-08-07": {{Threshold: 272000, ModelRatio: 1.25, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| "gpt-5-chat-latest": {{Threshold: 272000, ModelRatio: 1.25, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| "gpt-5-mini": {{Threshold: 272000, ModelRatio: 0.25, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| "gpt-5-mini-2025-08-07": {{Threshold: 272000, ModelRatio: 0.25, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| "gpt-5-nano": {{Threshold: 272000, ModelRatio: 0.05, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| "gpt-5-nano-2025-08-07": {{Threshold: 272000, ModelRatio: 0.05, CompletionRatio: 6.0, CacheRatio: 0.2}}, | ||
| } | ||
|
|
||
| var tieredPricingMap = types.NewRWMap[string, []TieredPricingTier]() | ||
|
|
||
| func TieredPricing2JSONString() string { | ||
| return tieredPricingMap.MarshalJSONString() | ||
| } | ||
|
|
||
| func UpdateTieredPricingByJSONString(jsonStr string) error { | ||
| return types.LoadFromJsonStringWithCallback(tieredPricingMap, jsonStr, InvalidateExposedDataCache) | ||
|
Comment on lines
+45
to
+46
Contributor
There was a problem hiding this comment. Normalize and validate tiers before publishing them.
🤖 Prompt for AI Agents |
||
| } | ||
|
|
||
| func GetTieredPricingCopy() map[string][]TieredPricingTier { | ||
| return tieredPricingMap.ReadAll() | ||
| } | ||
|
|
||
| // GetTieredPricing returns the tiered pricing tiers for a model. | ||
| // Returns nil if the model has no tiered pricing. | ||
| func GetTieredPricing(name string) []TieredPricingTier { | ||
| name = FormatMatchingModelName(name) | ||
| tiers, ok := tieredPricingMap.Get(name) | ||
| if !ok || len(tiers) == 0 { | ||
| return nil | ||
| } | ||
| return tiers | ||
| } | ||
|
|
||
| // ResolveTieredPricing determines the active tier based on prompt-side token count. | ||
| // Returns the tier if prompt tokens exceed a threshold, nil otherwise (use base ratios). | ||
| func ResolveTieredPricing(name string, promptSideTokens int) *TieredPricingTier { | ||
| tiers := GetTieredPricing(name) | ||
| if tiers == nil { | ||
| return nil | ||
| } | ||
|
|
||
| // Find the highest tier whose threshold is <= promptSideTokens | ||
| var activeTier *TieredPricingTier | ||
| for i := range tiers { | ||
| if promptSideTokens >= tiers[i].Threshold { | ||
| activeTier = &tiers[i] | ||
| } else { | ||
| break | ||
| } | ||
| } | ||
| return activeTier | ||
| } | ||
|
|
||
| // GetHighestTier returns the highest (most expensive) tier for a model. | ||
| // Used for conservative pre-consumption estimation. | ||
| func GetHighestTier(name string) *TieredPricingTier { | ||
| tiers := GetTieredPricing(name) | ||
| if tiers == nil { | ||
| return nil | ||
| } | ||
| return &tiers[len(tiers)-1] | ||
| } | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Don't base free-model detection on the non-tiered ratio.
This block can pre-consume using
highestTier.ModelRatio, but Line 125 still decides "free model" from the basemodelRatio. If a model is free below the threshold and paid above it, paid tiered requests will have their pre-consumption zeroed back out.[suggested fix: track the effective ratio used for pre-consumption and reuse that in the later free-model check.]
💡 Proposed fix
🤖 Prompt for AI Agents