diff --git a/relay/billing/ratio/model.go b/relay/billing/ratio/model.go index ef7898a8..617a9c76 100644 --- a/relay/billing/ratio/model.go +++ b/relay/billing/ratio/model.go @@ -272,21 +272,21 @@ var ModelRatio = map[string]float64{ "deepseek-r1-distill-qwen-1.5b": 0.001 * KiloRmb, "deepseek-r1-distill-qwen-7b": 0.0005 * KiloRmb, "deepseek-r1-distill-qwen-14b": 0.001 * KiloRmb, - "deepseek-r1-distill-qwen-32b": 0.002 * KiloRmb, - "deepseek-r1-distill-llama-8b": 0.0005 * KiloRmb, - "deepseek-r1-distill-llama-70b": 0.004 * KiloRmb, - "SparkDesk": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v1.1": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v2.1": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v3.1": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v3.1-128K": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v3.5": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v3.5-32K": 1.2858, // ￥0.018 / 1k tokens - "SparkDesk-v4.0": 1.2858, // ￥0.018 / 1k tokens - "360GPT_S2_V9": 0.8572, // ¥0.012 / 1k tokens - "embedding-bert-512-v1": 0.0715, // ¥0.001 / 1k tokens - "embedding_s1_v1": 0.0715, // ¥0.001 / 1k tokens - "semantic_similarity_s1_v1": 0.0715, // ¥0.001 / 1k tokens + // "deepseek-r1-distill-qwen-32b": 0.002 * KiloRmb, + "deepseek-r1-distill-llama-8b": 0.0005 * KiloRmb, + // "deepseek-r1-distill-llama-70b": 0.004 * KiloRmb, + "SparkDesk": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v1.1": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v2.1": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v3.1": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v3.1-128K": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v3.5": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v3.5-32K": 1.2858, // ￥0.018 / 1k tokens + "SparkDesk-v4.0": 1.2858, // ￥0.018 / 1k tokens + "360GPT_S2_V9": 0.8572, // ¥0.012 / 1k tokens + "embedding-bert-512-v1": 0.0715, // ¥0.001 / 1k tokens + "embedding_s1_v1": 0.0715, // ¥0.001 / 1k tokens + "semantic_similarity_s1_v1": 0.0715, // ¥0.001 / 1k tokens // https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0 "hunyuan-turbo": 0.015 * KiloRmb, "hunyuan-large": 0.004 * KiloRmb, @@ -321,22 +321,30 @@ var ModelRatio = map[string]float64{ "mistral-medium-latest": 2.7 * MilliTokensUsd, "mistral-large-latest": 8.0 * MilliTokensUsd, "mistral-embed": 0.1 * MilliTokensUsd, - // https://wow.groq.com/#:~:text=inquiries%C2%A0here.-,Model,-Current%20Speed - "gemma-7b-it": 0.07 / 1000 * MilliTokensUsd, - "gemma2-9b-it": 0.20 / 1000 * MilliTokensUsd, - "llama-3.1-70b-versatile": 0.59 / 1000 * MilliTokensUsd, - "llama-3.1-8b-instant": 0.05 / 1000 * MilliTokensUsd, - "llama-3.2-11b-text-preview": 0.05 / 1000 * MilliTokensUsd, - "llama-3.2-11b-vision-preview": 0.05 / 1000 * MilliTokensUsd, - "llama-3.2-1b-preview": 0.05 / 1000 * MilliTokensUsd, - "llama-3.2-3b-preview": 0.05 / 1000 * MilliTokensUsd, - "llama-3.2-90b-text-preview": 0.59 / 1000 * MilliTokensUsd, - "llama-guard-3-8b": 0.05 / 1000 * MilliTokensUsd, - "llama3-70b-8192": 0.59 / 1000 * MilliTokensUsd, - "llama3-8b-8192": 0.05 / 1000 * MilliTokensUsd, - "llama3-groq-70b-8192-tool-use-preview": 0.89 / 1000 * MilliTokensUsd, - "llama3-groq-8b-8192-tool-use-preview": 0.19 / 1000 * MilliTokensUsd, - "mixtral-8x7b-32768": 0.24 / 1000 * MilliTokensUsd, + // ------------------------------------- + // https://groq.com/pricing/ + // ------------------------------------- + "gemma2-9b-it": 0.20 * MilliTokensUsd, + "llama-3.1-8b-instant": 0.05 * MilliTokensUsd, + "llama-3.2-11b-text-preview": 0.18 * MilliTokensUsd, + "llama-3.2-11b-vision-preview": 0.18 * MilliTokensUsd, + "llama-3.2-1b-preview": 0.04 * MilliTokensUsd, + "llama-3.2-3b-preview": 0.06 * MilliTokensUsd, + "llama-3.2-90b-text-preview": 0.90 * MilliTokensUsd, + "llama-3.2-90b-vision-preview": 0.90 * MilliTokensUsd, + "llama-3.3-70b-versatile": 0.59 * MilliTokensUsd, + "llama-guard-3-8b": 0.20 * MilliTokensUsd, + "llama3-70b-8192": 0.59 * MilliTokensUsd, + "llama3-8b-8192": 0.05 * MilliTokensUsd, + "llama3-groq-70b-8192-tool-use-preview": 0.59 * MilliTokensUsd, + "llama3-groq-8b-8192-tool-use-preview": 0.05 * MilliTokensUsd, + "mixtral-8x7b-32768": 0.24 * MilliTokensUsd, + "whisper-large-v3": 0.111 * MilliTokensUsd, + "whisper-large-v3-turbo": 0.04 * MilliTokensUsd, + "distil-whisper-large-v3-en": 0.02 * MilliTokensUsd, + "deepseek-r1-distill-qwen-32b": 0.69 * MilliTokensUsd, + "deepseek-r1-distill-llama-70b-specdec": 0.75 * MilliTokensUsd, + "deepseek-r1-distill-llama-70b": 0.75 * MilliTokensUsd, // https://platform.lingyiwanwu.com/docs#-计费单元 "yi-34b-chat-0205": 2.5 * MilliRmb, "yi-34b-chat-200k": 12.0 * MilliRmb, @@ -651,15 +659,113 @@ var ModelRatio = map[string]float64{ "xwin-lm/xwin-lm-70b": 1.875, } +// CompletionRatio is the price ratio between completion tokens and prompt tokens var CompletionRatio = map[string]float64{ // aws llama3 "llama3-8b-8192(33)": 0.0006 / 0.0003, "llama3-70b-8192(33)": 0.0035 / 0.00265, // whisper - "whisper-1": 0, // only count input tokens + "whisper-1": 0, // only count input tokens + "whisper-large-v3": 0, // only count input tokens + "whisper-large-v3-turbo": 0, // only count input tokens + "distil-whisper-large-v3-en": 0, // only count input tokens // deepseek - "deepseek-chat": 0.28 / 0.14, + "deepseek-chat": 1.1 / 0.27, "deepseek-reasoner": 2.19 / 0.55, + // openrouter + "deepseek/deepseek-chat": 1, + "deepseek/deepseek-r1": 1, + // ------------------------------------- + // groq + // ------------------------------------- + "llama-3.3-70b-versatile": 0.79 / 0.59, + "llama-3.1-8b-instant": 0.08 / 0.05, + "llama3-70b-8192": 0.79 / 0.59, + "llama3-8b-8192": 0.08 / 0.05, + "gemma2-9b-it": 1.0, + "llama-3.2-11b-text-preview": 1.0, + "llama-3.2-11b-vision-preview": 1.0, + "llama-3.2-1b-preview": 1.0, + "llama-3.2-3b-preview": 1.0, + "llama-3.2-90b-text-preview": 1.0, + "llama-3.2-90b-vision-preview": 1.0, + "llama-guard-3-8b": 1.0, + "llama3-groq-70b-8192-tool-use-preview": 0.79 / 0.59, + "llama3-groq-8b-8192-tool-use-preview": 0.08 / 0.05, + "mixtral-8x7b-32768": 1.0, + "deepseek-r1-distill-qwen-32b": 1, + "deepseek-r1-distill-llama-70b-specdec": 0.99 / 0.75, + "deepseek-r1-distill-llama-70b": 0.99 / 0.75, +} + +// AudioRatio represents the price ratio between audio tokens and text tokens +var AudioRatio = map[string]float64{ + "gpt-4o-audio-preview": 16, + "gpt-4o-audio-preview-2024-12-17": 16, + "gpt-4o-audio-preview-2024-10-01": 40, + "gpt-4o-mini-audio-preview": 10 / 0.15, + "gpt-4o-mini-audio-preview-2024-12-17": 10 / 0.15, +} + +// GetAudioPromptRatio returns the audio prompt ratio for the given model. +func GetAudioPromptRatio(actualModelName string) float64 { + var v float64 + if ratio, ok := AudioRatio[actualModelName]; ok { + v = ratio + } else { + v = 16 + } + + return v +} + +// AudioCompletionRatio is the completion ratio for audio models. +var AudioCompletionRatio = map[string]float64{ + "whisper-1": 0, + "gpt-4o-audio-preview": 2, + "gpt-4o-audio-preview-2024-12-17": 2, + "gpt-4o-audio-preview-2024-10-01": 2, + "gpt-4o-mini-audio-preview": 2, + "gpt-4o-mini-audio-preview-2024-12-17": 2, +} + +// GetAudioCompletionRatio returns the completion ratio for audio models. +func GetAudioCompletionRatio(actualModelName string) float64 { + var v float64 + if ratio, ok := AudioCompletionRatio[actualModelName]; ok { + v = ratio + } else { + v = 2 + } + + return v +} + +// AudioTokensPerSecond is the number of audio tokens per second for each model. +var AudioPromptTokensPerSecond = map[string]float64{ + // Whisper API price is $0.0001/sec. One-api's historical ratio is 15, + // corresponding to $0.03/kilo_tokens. + // After conversion, tokens per second should be 0.0001/0.03*1000 = 3.3333. + "whisper-1": 0.0001 / 0.03 * 1000, + // gpt-4o-audio series processes 10 tokens per second + "gpt-4o-audio-preview": 10, + "gpt-4o-audio-preview-2024-12-17": 10, + "gpt-4o-audio-preview-2024-10-01": 10, + "gpt-4o-mini-audio-preview": 10, + "gpt-4o-mini-audio-preview-2024-12-17": 10, +} + +// GetAudioPromptTokensPerSecond returns the number of audio tokens per second +// for the given model. +func GetAudioPromptTokensPerSecond(actualModelName string) float64 { + var v float64 + if tokensPerSecond, ok := AudioPromptTokensPerSecond[actualModelName]; ok { + v = tokensPerSecond + } else { + v = 10 + } + + return v } var ( @@ -722,23 +828,26 @@ func GetModelRatio(name string, channelType int) float64 { if strings.HasPrefix(name, "command-") && strings.HasSuffix(name, "-internet") { name = strings.TrimSuffix(name, "-internet") } + model := fmt.Sprintf("%s(%d)", name, channelType) - if ratio, ok := ModelRatio[model]; ok { - return ratio - } - if ratio, ok := DefaultModelRatio[model]; ok { - return ratio - } - if ratio, ok := ModelRatio[name]; ok { - return ratio - } - if ratio, ok := DefaultModelRatio[name]; ok { - return ratio + + for _, targetName := range []string{model, name} { + for _, ratioMap := range []map[string]float64{ + ModelRatio, + DefaultModelRatio, + AudioRatio, + } { + if ratio, ok := ratioMap[targetName]; ok { + return ratio + } + } } + logger.SysError("model ratio not found: " + name) return 30 } +// CompletionRatio2JSONString returns the CompletionRatio map as a JSON string. func CompletionRatio2JSONString() string { jsonBytes, err := json.Marshal(CompletionRatio) if err != nil { @@ -747,59 +856,79 @@ func CompletionRatio2JSONString() string { return string(jsonBytes) } +// completionRatioLock is a mutex for synchronizing access to the CompletionRatio map. +var completionRatioLock sync.RWMutex + +// UpdateCompletionRatioByJSONString updates the CompletionRatio map with the given JSON string. func UpdateCompletionRatioByJSONString(jsonStr string) error { + completionRatioLock.Lock() + defer completionRatioLock.Unlock() CompletionRatio = make(map[string]float64) return json.Unmarshal([]byte(jsonStr), &CompletionRatio) } +// GetCompletionRatio returns the completion ratio for the given model name and channel type. func GetCompletionRatio(name string, channelType int) float64 { + completionRatioLock.RLock() + defer completionRatioLock.RUnlock() if strings.HasPrefix(name, "qwen-") && strings.HasSuffix(name, "-internet") { name = strings.TrimSuffix(name, "-internet") } model := fmt.Sprintf("%s(%d)", name, channelType) - if ratio, ok := CompletionRatio[model]; ok { - return ratio + + name = strings.TrimPrefix(name, "openai/") + for _, targetName := range []string{model, name} { + for _, ratioMap := range []map[string]float64{ + CompletionRatio, + DefaultCompletionRatio, + AudioCompletionRatio, + } { + // first try the model name + if ratio, ok := ratioMap[targetName]; ok { + return ratio + } + + // then try the model name without some special prefix + normalizedTargetName := strings.TrimPrefix(targetName, "openai/") + if ratio, ok := ratioMap[normalizedTargetName]; ok { + return ratio + } + } } - if ratio, ok := DefaultCompletionRatio[model]; ok { - return ratio - } - if ratio, ok := CompletionRatio[name]; ok { - return ratio - } - if ratio, ok := DefaultCompletionRatio[name]; ok { - return ratio - } - if strings.HasPrefix(name, "gpt-3.5") { - if name == "gpt-3.5-turbo" || strings.HasSuffix(name, "0125") { + + // openai + switch { + case strings.HasPrefix(name, "gpt-3.5"): + switch { + case name == "gpt-3.5-turbo" || strings.HasSuffix(name, "0125"): // https://openai.com/blog/new-embedding-models-and-api-updates // Updated GPT-3.5 Turbo model and lower pricing return 3 - } - if strings.HasSuffix(name, "1106") { + case strings.HasSuffix(name, "1106"): return 2 + default: + return 4.0 / 3.0 } - return 4.0 / 3.0 - } - if strings.HasPrefix(name, "gpt-4") { - if strings.HasPrefix(name, "gpt-4o") { + case name == "chatgpt-4o-latest": + return 3 + case strings.HasPrefix(name, "gpt-4"): + switch { + case strings.HasPrefix(name, "gpt-4o"): if name == "gpt-4o-2024-05-13" { return 3 } return 4 - } - if strings.HasPrefix(name, "gpt-4-turbo") || - strings.HasSuffix(name, "preview") { + case strings.HasPrefix(name, "gpt-4-"): return 3 + default: + return 2 } - return 2 - } - // including o1, o1-preview, o1-mini - if strings.HasPrefix(name, "o1") { + // including o1/o1-preview/o1-mini + case strings.HasPrefix(name, "o1") || + strings.HasPrefix(name, "o3"): return 4 } - if name == "chatgpt-4o-latest" { - return 3 - } + if strings.HasPrefix(name, "claude-3") { return 5 } @@ -810,10 +939,7 @@ func GetCompletionRatio(name string, channelType int) float64 { return 3 } if strings.HasPrefix(name, "gemini-") { - return 3 - } - if strings.HasPrefix(name, "deepseek-") { - return 2 + return 4 } switch name { @@ -861,5 +987,6 @@ func GetCompletionRatio(name string, channelType int) float64 { return 1.000 / 0.300 // ≈3.333333 } + logger.SysWarn(fmt.Sprintf("completion ratio not found for model: %s (channel type: %d), using default value 1", name, channelType)) return 1 }