mirror of
https://github.com/songquanpeng/one-api.git
synced 2025-09-29 14:46:38 +08:00
338 lines
14 KiB
Go
338 lines
14 KiB
Go
package common
|
||
|
||
import (
|
||
"encoding/json"
|
||
"strings"
|
||
"time"
|
||
)
|
||
|
||
type ModelType struct {
|
||
Ratio []float64
|
||
Type int
|
||
}
|
||
|
||
var ModelTypes map[string]ModelType
|
||
|
||
// ModelRatio
|
||
// https://platform.openai.com/docs/models/model-endpoint-compatibility
|
||
// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/Blfmc9dlf
|
||
// https://openai.com/pricing
|
||
// TODO: when a new api is enabled, check the pricing here
|
||
// 1 === $0.002 / 1K tokens
|
||
// 1 === ￥0.014 / 1k tokens
|
||
var ModelRatio map[string][]float64
|
||
|
||
func init() {
|
||
ModelTypes = map[string]ModelType{
|
||
// $0.03 / 1K tokens $0.06 / 1K tokens
|
||
"gpt-4": {[]float64{15, 30}, ChannelTypeOpenAI},
|
||
"gpt-4-0314": {[]float64{15, 30}, ChannelTypeOpenAI},
|
||
"gpt-4-0613": {[]float64{15, 30}, ChannelTypeOpenAI},
|
||
// $0.06 / 1K tokens $0.12 / 1K tokens
|
||
"gpt-4-32k": {[]float64{30, 60}, ChannelTypeOpenAI},
|
||
"gpt-4-32k-0314": {[]float64{30, 60}, ChannelTypeOpenAI},
|
||
"gpt-4-32k-0613": {[]float64{30, 60}, ChannelTypeOpenAI},
|
||
// $0.01 / 1K tokens $0.03 / 1K tokens
|
||
"gpt-4-preview": {[]float64{5, 15}, ChannelTypeOpenAI},
|
||
"gpt-4-1106-preview": {[]float64{5, 15}, ChannelTypeOpenAI},
|
||
"gpt-4-0125-preview": {[]float64{5, 15}, ChannelTypeOpenAI},
|
||
"gpt-4-turbo-preview": {[]float64{5, 15}, ChannelTypeOpenAI},
|
||
"gpt-4-vision-preview": {[]float64{5, 15}, ChannelTypeOpenAI},
|
||
// $0.0005 / 1K tokens $0.0015 / 1K tokens
|
||
"gpt-3.5-turbo": {[]float64{0.25, 0.75}, ChannelTypeOpenAI},
|
||
"gpt-3.5-turbo-0125": {[]float64{0.25, 0.75}, ChannelTypeOpenAI},
|
||
// $0.0015 / 1K tokens $0.002 / 1K tokens
|
||
"gpt-3.5-turbo-0301": {[]float64{0.75, 1}, ChannelTypeOpenAI},
|
||
"gpt-3.5-turbo-0613": {[]float64{0.75, 1}, ChannelTypeOpenAI},
|
||
"gpt-3.5-turbo-instruct": {[]float64{0.75, 1}, ChannelTypeOpenAI},
|
||
// $0.003 / 1K tokens $0.004 / 1K tokens
|
||
"gpt-3.5-turbo-16k": {[]float64{1.5, 2}, ChannelTypeOpenAI},
|
||
"gpt-3.5-turbo-16k-0613": {[]float64{1.5, 2}, ChannelTypeOpenAI},
|
||
// $0.001 / 1K tokens $0.002 / 1K tokens
|
||
"gpt-3.5-turbo-1106": {[]float64{0.5, 1}, ChannelTypeOpenAI},
|
||
// $0.0020 / 1K tokens
|
||
"davinci-002": {[]float64{1, 1}, ChannelTypeOpenAI},
|
||
// $0.0004 / 1K tokens
|
||
"babbage-002": {[]float64{0.2, 0.2}, ChannelTypeOpenAI},
|
||
"text-ada-001": {[]float64{0.2, 0.2}, ChannelTypeOpenAI},
|
||
"text-babbage-001": {[]float64{0.25, 0.25}, ChannelTypeOpenAI},
|
||
"text-curie-001": {[]float64{1, 1}, ChannelTypeOpenAI},
|
||
"text-davinci-002": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"text-davinci-003": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"text-davinci-edit-001": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"code-davinci-edit-001": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
// $0.006 / minute -> $0.006 / 150 words -> $0.006 / 200 tokens -> $0.03 / 1k tokens
|
||
"whisper-1": {[]float64{15, 15}, ChannelTypeOpenAI},
|
||
// $0.015 / 1K characters
|
||
"tts-1": {[]float64{7.5, 7.5}, ChannelTypeOpenAI},
|
||
"tts-1-1106": {[]float64{7.5, 7.5}, ChannelTypeOpenAI},
|
||
// $0.030 / 1K characters
|
||
"tts-1-hd": {[]float64{15, 15}, ChannelTypeOpenAI},
|
||
"tts-1-hd-1106": {[]float64{15, 15}, ChannelTypeOpenAI},
|
||
"davinci": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"curie": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"babbage": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"ada": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"text-embedding-ada-002": {[]float64{0.05, 0.05}, ChannelTypeOpenAI},
|
||
// $0.00002 / 1K tokens
|
||
"text-embedding-3-small": {[]float64{0.01, 0.01}, ChannelTypeOpenAI},
|
||
// $0.00013 / 1K tokens
|
||
"text-embedding-3-large": {[]float64{0.065, 0.065}, ChannelTypeOpenAI},
|
||
"text-search-ada-doc-001": {[]float64{10, 10}, ChannelTypeOpenAI},
|
||
"text-moderation-stable": {[]float64{0.1, 0.1}, ChannelTypeOpenAI},
|
||
"text-moderation-latest": {[]float64{0.1, 0.1}, ChannelTypeOpenAI},
|
||
// $0.016 - $0.020 / image
|
||
"dall-e-2": {[]float64{8, 8}, ChannelTypeOpenAI},
|
||
// $0.040 - $0.120 / image
|
||
"dall-e-3": {[]float64{20, 20}, ChannelTypeOpenAI},
|
||
|
||
// $0.80/million tokens $2.40/million tokens
|
||
"claude-instant-1.2": {[]float64{0.4, 1.2}, ChannelTypeAnthropic},
|
||
// $8.00/million tokens $24.00/million tokens
|
||
"claude-2.0": {[]float64{4, 12}, ChannelTypeAnthropic},
|
||
"claude-2.1": {[]float64{4, 12}, ChannelTypeAnthropic},
|
||
// $15 / M $75 / M
|
||
"claude-3-opus-20240229": {[]float64{7.5, 22.5}, ChannelTypeAnthropic},
|
||
// $3 / M $15 / M
|
||
"claude-3-sonnet-20240229": {[]float64{1.3, 3.9}, ChannelTypeAnthropic},
|
||
// $0.25 / M $1.25 / M 0.00025$ / 1k tokens 0.00125$ / 1k tokens
|
||
"claude-3-haiku-20240307": {[]float64{0.125, 0.625}, ChannelTypeAnthropic},
|
||
|
||
// ￥0.004 / 1k tokens ￥0.008 / 1k tokens
|
||
"ERNIE-Speed": {[]float64{0.2857, 0.5714}, ChannelTypeBaidu},
|
||
// ￥0.012 / 1k tokens ￥0.012 / 1k tokens
|
||
"ERNIE-Bot": {[]float64{0.8572, 0.8572}, ChannelTypeBaidu},
|
||
"ERNIE-3.5-8K": {[]float64{0.8572, 0.8572}, ChannelTypeBaidu},
|
||
// 0.024元/千tokens 0.048元/千tokens
|
||
"ERNIE-Bot-8k": {[]float64{1.7143, 3.4286}, ChannelTypeBaidu},
|
||
// ￥0.008 / 1k tokens ￥0.008 / 1k tokens
|
||
"ERNIE-Bot-turbo": {[]float64{0.5715, 0.5715}, ChannelTypeBaidu},
|
||
// ￥0.12 / 1k tokens ￥0.12 / 1k tokens
|
||
"ERNIE-Bot-4": {[]float64{8.572, 8.572}, ChannelTypeBaidu},
|
||
"ERNIE-4.0": {[]float64{8.572, 8.572}, ChannelTypeBaidu},
|
||
// ￥0.002 / 1k tokens
|
||
"Embedding-V1": {[]float64{0.1429, 0.1429}, ChannelTypeBaidu},
|
||
// ￥0.004 / 1k tokens
|
||
"BLOOMZ-7B": {[]float64{0.2857, 0.2857}, ChannelTypeBaidu},
|
||
|
||
"PaLM-2": {[]float64{1, 1}, ChannelTypePaLM},
|
||
"gemini-pro": {[]float64{1, 1}, ChannelTypeGemini},
|
||
"gemini-pro-vision": {[]float64{1, 1}, ChannelTypeGemini},
|
||
"gemini-1.0-pro": {[]float64{1, 1}, ChannelTypeGemini},
|
||
"gemini-1.5-pro": {[]float64{1, 1}, ChannelTypeGemini},
|
||
|
||
// ￥0.005 / 1k tokens
|
||
"glm-3-turbo": {[]float64{0.3572, 0.3572}, ChannelTypeZhipu},
|
||
// ￥0.1 / 1k tokens
|
||
"glm-4": {[]float64{7.143, 7.143}, ChannelTypeZhipu},
|
||
"glm-4v": {[]float64{7.143, 7.143}, ChannelTypeZhipu},
|
||
// ￥0.0005 / 1k tokens
|
||
"embedding-2": {[]float64{0.0357, 0.0357}, ChannelTypeZhipu},
|
||
// ￥0.25 / 1张图片
|
||
"cogview-3": {[]float64{17.8571, 17.8571}, ChannelTypeZhipu},
|
||
|
||
// ￥0.008 / 1k tokens
|
||
"qwen-turbo": {[]float64{0.5715, 0.5715}, ChannelTypeAli},
|
||
// ￥0.02 / 1k tokens
|
||
"qwen-plus": {[]float64{1.4286, 1.4286}, ChannelTypeAli},
|
||
"qwen-vl-max": {[]float64{1.4286, 1.4286}, ChannelTypeAli},
|
||
// 0.12元/1,000tokens
|
||
"qwen-max": {[]float64{8.5714, 8.5714}, ChannelTypeAli},
|
||
"qwen-max-longcontext": {[]float64{8.5714, 8.5714}, ChannelTypeAli},
|
||
// 0.008元/1,000tokens
|
||
"qwen-vl-plus": {[]float64{0.5715, 0.5715}, ChannelTypeAli},
|
||
// ￥0.0007 / 1k tokens
|
||
"text-embedding-v1": {[]float64{0.05, 0.05}, ChannelTypeAli},
|
||
|
||
// ￥0.018 / 1k tokens
|
||
"SparkDesk": {[]float64{1.2858, 1.2858}, ChannelTypeXunfei},
|
||
"SparkDesk-v1.1": {[]float64{1.2858, 1.2858}, ChannelTypeXunfei},
|
||
"SparkDesk-v2.1": {[]float64{1.2858, 1.2858}, ChannelTypeXunfei},
|
||
"SparkDesk-v3.1": {[]float64{1.2858, 1.2858}, ChannelTypeXunfei},
|
||
"SparkDesk-v3.5": {[]float64{1.2858, 1.2858}, ChannelTypeXunfei},
|
||
|
||
// ¥0.012 / 1k tokens
|
||
"360GPT_S2_V9": {[]float64{0.8572, 0.8572}, ChannelType360},
|
||
// ¥0.001 / 1k tokens
|
||
"embedding-bert-512-v1": {[]float64{0.0715, 0.0715}, ChannelType360},
|
||
"embedding_s1_v1": {[]float64{0.0715, 0.0715}, ChannelType360},
|
||
"semantic_similarity_s1_v1": {[]float64{0.0715, 0.0715}, ChannelType360},
|
||
|
||
// ¥0.1 / 1k tokens // https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0
|
||
"hunyuan": {[]float64{7.143, 7.143}, ChannelTypeTencent},
|
||
// https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0
|
||
// ¥0.01 / 1k tokens
|
||
"ChatStd": {[]float64{0.7143, 0.7143}, ChannelTypeTencent},
|
||
//¥0.1 / 1k tokens
|
||
"ChatPro": {[]float64{7.143, 7.143}, ChannelTypeTencent},
|
||
|
||
"Baichuan2-Turbo": {[]float64{0.5715, 0.5715}, ChannelTypeBaichuan}, // ¥0.008 / 1k tokens
|
||
"Baichuan2-Turbo-192k": {[]float64{1.143, 1.143}, ChannelTypeBaichuan}, // ¥0.016 / 1k tokens
|
||
"Baichuan2-53B": {[]float64{1.4286, 1.4286}, ChannelTypeBaichuan}, // ¥0.02 / 1k tokens
|
||
"Baichuan-Text-Embedding": {[]float64{0.0357, 0.0357}, ChannelTypeBaichuan}, // ¥0.0005 / 1k tokens
|
||
|
||
"abab5.5s-chat": {[]float64{0.3572, 0.3572}, ChannelTypeMiniMax}, // ¥0.005 / 1k tokens
|
||
"abab5.5-chat": {[]float64{1.0714, 1.0714}, ChannelTypeMiniMax}, // ¥0.015 / 1k tokens
|
||
"abab6-chat": {[]float64{14.2857, 14.2857}, ChannelTypeMiniMax}, // ¥0.2 / 1k tokens
|
||
"embo-01": {[]float64{0.0357, 0.0357}, ChannelTypeMiniMax}, // ¥0.0005 / 1k tokens
|
||
|
||
"deepseek-coder": {[]float64{0.75, 0.75}, ChannelTypeDeepseek}, // 暂定 $0.0015 / 1K tokens
|
||
"deepseek-chat": {[]float64{0.75, 0.75}, ChannelTypeDeepseek}, // 暂定 $0.0015 / 1K tokens
|
||
|
||
"moonshot-v1-8k": {[]float64{0.8572, 0.8572}, ChannelTypeMoonshot}, // ¥0.012 / 1K tokens
|
||
"moonshot-v1-32k": {[]float64{1.7143, 1.7143}, ChannelTypeMoonshot}, // ¥0.024 / 1K tokens
|
||
"moonshot-v1-128k": {[]float64{4.2857, 4.2857}, ChannelTypeMoonshot}, // ¥0.06 / 1K tokens
|
||
|
||
"open-mistral-7b": {[]float64{0.125, 0.125}, ChannelTypeMistral}, // 0.25$ / 1M tokens 0.25$ / 1M tokens 0.00025$ / 1k tokens
|
||
"open-mixtral-8x7b": {[]float64{0.35, 0.35}, ChannelTypeMistral}, // 0.7$ / 1M tokens 0.7$ / 1M tokens 0.0007$ / 1k tokens
|
||
"mistral-small-latest": {[]float64{1, 3}, ChannelTypeMistral}, // 2$ / 1M tokens 6$ / 1M tokens 0.002$ / 1k tokens
|
||
"mistral-medium-latest": {[]float64{1.35, 4.05}, ChannelTypeMistral}, // 2.7$ / 1M tokens 8.1$ / 1M tokens 0.0027$ / 1k tokens
|
||
"mistral-large-latest": {[]float64{4, 12}, ChannelTypeMistral}, // 8$ / 1M tokens 24$ / 1M tokens 0.008$ / 1k tokens
|
||
"mistral-embed": {[]float64{0.05, 0.05}, ChannelTypeMistral}, // 0.1$ / 1M tokens 0.1$ / 1M tokens 0.0001$ / 1k tokens
|
||
|
||
// $0.70/$0.80 /1M Tokens 0.0007$ / 1k tokens
|
||
"llama2-70b-4096": {[]float64{0.35, 0.4}, ChannelTypeGroq},
|
||
// $0.10/$0.10 /1M Tokens 0.0001$ / 1k tokens
|
||
"llama2-7b-2048": {[]float64{0.05, 0.05}, ChannelTypeGroq},
|
||
"gemma-7b-it": {[]float64{0.05, 0.05}, ChannelTypeGroq},
|
||
// $0.27/$0.27 /1M Tokens 0.00027$ / 1k tokens
|
||
"mixtral-8x7b-32768": {[]float64{0.135, 0.135}, ChannelTypeGroq},
|
||
|
||
// 2.5 元 / 1M tokens 0.0025 / 1k tokens
|
||
"yi-34b-chat-0205": {[]float64{0.1786, 0.1786}, ChannelTypeLingyi},
|
||
// 12 元 / 1M tokens 0.012 / 1k tokens
|
||
"yi-34b-chat-200k": {[]float64{0.8571, 0.8571}, ChannelTypeLingyi},
|
||
// 6 元 / 1M tokens 0.006 / 1k tokens
|
||
"yi-vl-plus": {[]float64{0.4286, 0.4286}, ChannelTypeLingyi},
|
||
}
|
||
|
||
ModelRatio = make(map[string][]float64)
|
||
for name, modelType := range ModelTypes {
|
||
ModelRatio[name] = modelType.Ratio
|
||
}
|
||
}
|
||
|
||
var DalleSizeRatios = map[string]map[string]float64{
|
||
"dall-e-2": {
|
||
"256x256": 1,
|
||
"512x512": 1.125,
|
||
"1024x1024": 1.25,
|
||
},
|
||
"dall-e-3": {
|
||
"1024x1024": 1,
|
||
"1024x1792": 2,
|
||
"1792x1024": 2,
|
||
},
|
||
}
|
||
|
||
var DalleGenerationImageAmounts = map[string][2]int{
|
||
"dall-e-2": {1, 10},
|
||
"dall-e-3": {1, 1}, // OpenAI allows n=1 currently.
|
||
}
|
||
|
||
var DalleImagePromptLengthLimitations = map[string]int{
|
||
"dall-e-2": 1000,
|
||
"dall-e-3": 4000,
|
||
}
|
||
|
||
func ModelRatio2JSONString() string {
|
||
jsonBytes, err := json.Marshal(ModelRatio)
|
||
if err != nil {
|
||
SysError("error marshalling model ratio: " + err.Error())
|
||
}
|
||
return string(jsonBytes)
|
||
}
|
||
|
||
func UpdateModelRatioByJSONString(jsonStr string) error {
|
||
ModelRatio = make(map[string][]float64)
|
||
return json.Unmarshal([]byte(jsonStr), &ModelRatio)
|
||
}
|
||
|
||
func MergeModelRatioByJSONString(jsonStr string) (newJsonStr string, err error) {
|
||
isNew := false
|
||
inputModelRatio := make(map[string][]float64)
|
||
err = json.Unmarshal([]byte(jsonStr), &inputModelRatio)
|
||
if err != nil {
|
||
inputModelRatioOld := make(map[string]float64)
|
||
err = json.Unmarshal([]byte(jsonStr), &inputModelRatioOld)
|
||
if err != nil {
|
||
return
|
||
}
|
||
|
||
inputModelRatio = UpdateModeRatioFormat(inputModelRatioOld)
|
||
isNew = true
|
||
}
|
||
|
||
// 与现有的ModelRatio进行比较，如果有新增的模型，需要添加
|
||
for key, value := range ModelRatio {
|
||
if _, ok := inputModelRatio[key]; !ok {
|
||
isNew = true
|
||
inputModelRatio[key] = value
|
||
}
|
||
}
|
||
|
||
if !isNew {
|
||
return
|
||
}
|
||
|
||
var jsonBytes []byte
|
||
jsonBytes, err = json.Marshal(inputModelRatio)
|
||
if err != nil {
|
||
SysError("error marshalling model ratio: " + err.Error())
|
||
}
|
||
newJsonStr = string(jsonBytes)
|
||
return
|
||
}
|
||
|
||
func UpdateModeRatioFormat(modelRatioOld map[string]float64) map[string][]float64 {
|
||
modelRatioNew := make(map[string][]float64)
|
||
for key, value := range modelRatioOld {
|
||
completionRatio := GetCompletionRatio(key) * value
|
||
modelRatioNew[key] = []float64{value, completionRatio}
|
||
}
|
||
return modelRatioNew
|
||
}
|
||
|
||
func GetModelRatio(name string) []float64 {
|
||
if strings.HasPrefix(name, "qwen-") && strings.HasSuffix(name, "-internet") {
|
||
name = strings.TrimSuffix(name, "-internet")
|
||
}
|
||
ratio, ok := ModelRatio[name]
|
||
if !ok {
|
||
SysError("model ratio not found: " + name)
|
||
return []float64{30, 30}
|
||
}
|
||
return ratio
|
||
}
|
||
|
||
func GetCompletionRatio(name string) float64 {
|
||
if strings.HasPrefix(name, "gpt-3.5") {
|
||
if strings.HasSuffix(name, "1106") {
|
||
return 2
|
||
}
|
||
if name == "gpt-3.5-turbo" || name == "gpt-3.5-turbo-16k" {
|
||
// TODO: clear this after 2023-12-11
|
||
now := time.Now()
|
||
// https://platform.openai.com/docs/models/continuous-model-upgrades
|
||
// if after 2023-12-11, use 2
|
||
if now.After(time.Date(2023, 12, 11, 0, 0, 0, 0, time.UTC)) {
|
||
return 2
|
||
}
|
||
}
|
||
return 1.333333
|
||
}
|
||
if strings.HasPrefix(name, "gpt-4") {
|
||
if strings.HasSuffix(name, "preview") {
|
||
return 3
|
||
}
|
||
return 2
|
||
}
|
||
if strings.HasPrefix(name, "claude-instant-1.2") {
|
||
return 3.38
|
||
}
|
||
if strings.HasPrefix(name, "claude-2") {
|
||
return 2.965517
|
||
}
|
||
return 1
|
||
}
|