phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity

Update pricing and add cohere refresh models

Browse source
This commit is contained in:
Jay Alammar 2024-09-06 17:35:59 -04:00
parent 4626c5a365
commit 6d9f6a1b76
1 changed files with 633 additions and 633 deletions
Download patch file Download diff file Expand all files Collapse all files

132
model_prices_and_context_window.json
View file

@ -890,7 +890,7 @@
"output_cost_per_token": 0.00000061, "output_cost_per_token": 0.00000061,
"litellm_provider": "azure_ai", "litellm_provider": "azure_ai",
"mode": "chat", "mode": "chat",
"source":"https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-1-8b-instruct-offer?tab=PlansAndPrice" "source": "https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-1-8b-instruct-offer?tab=PlansAndPrice"
}, },
"azure_ai/Meta-Llama-31-70B-Instruct": { "azure_ai/Meta-Llama-31-70B-Instruct": {
"max_tokens": 128000, "max_tokens": 128000,
@ -900,7 +900,7 @@
"output_cost_per_token": 0.00000354, "output_cost_per_token": 0.00000354,
"litellm_provider": "azure_ai", "litellm_provider": "azure_ai",
"mode": "chat", "mode": "chat",
"source":"https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-1-70b-instruct-offer?tab=PlansAndPrice" "source": "https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-1-70b-instruct-offer?tab=PlansAndPrice"
}, },
"azure_ai/Meta-Llama-31-405B-Instruct": { "azure_ai/Meta-Llama-31-405B-Instruct": {
"max_tokens": 128000, "max_tokens": 128000,
@ -910,7 +910,7 @@
"output_cost_per_token": 0.000016, "output_cost_per_token": 0.000016,
"litellm_provider": "azure_ai", "litellm_provider": "azure_ai",
"mode": "chat", "mode": "chat",
"source":"https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-1-405b-instruct-offer?tab=PlansAndPrice" "source": "https://azuremarketplace.microsoft.com/en-us/marketplace/apps/metagenai.meta-llama-3-1-405b-instruct-offer?tab=PlansAndPrice"
}, },
"babbage-002": { "babbage-002": {
"max_tokens": 16384, "max_tokens": 16384,
@ -947,7 +947,6 @@
"output_cost_per_token": 0.000002, "output_cost_per_token": 0.000002,
"litellm_provider": "text-completion-openai", "litellm_provider": "text-completion-openai",
"mode": "completion" "mode": "completion"
}, },
"claude-instant-1": { "claude-instant-1": {
"max_tokens": 8191, "max_tokens": 8191,
@ -1109,7 +1108,7 @@
"max_tokens": 128000, "max_tokens": 128000,
"max_input_tokens": 128000, "max_input_tokens": 128000,
"max_output_tokens": 128000, "max_output_tokens": 128000,
"input_cost_per_token": 0.0000003, "input_cost_per_token": 0.0000003,
"output_cost_per_token": 0.0000003, "output_cost_per_token": 0.0000003,
"litellm_provider": "mistral", "litellm_provider": "mistral",
"mode": "chat", "mode": "chat",
@ -1120,7 +1119,7 @@
"max_tokens": 128000, "max_tokens": 128000,
"max_input_tokens": 128000, "max_input_tokens": 128000,
"max_output_tokens": 128000, "max_output_tokens": 128000,
"input_cost_per_token": 0.0000003, "input_cost_per_token": 0.0000003,
"output_cost_per_token": 0.0000003, "output_cost_per_token": 0.0000003,
"litellm_provider": "mistral", "litellm_provider": "mistral",
"mode": "chat", "mode": "chat",
@ -2432,7 +2431,7 @@
"mode": "embedding", "mode": "embedding",
"source": "https://cloud.google.com/vertex-ai/generative-ai/pricing" "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing"
}, },
"text-multilingual-embedding-preview-0409":{ "text-multilingual-embedding-preview-0409": {
"max_tokens": 3072, "max_tokens": 3072,
"max_input_tokens": 3072, "max_input_tokens": 3072,
"output_vector_size": 768, "output_vector_size": 768,
@ -2665,8 +2664,18 @@
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 128000, "max_input_tokens": 128000,
"max_output_tokens": 4096, "max_output_tokens": 4096,
"input_cost_per_token": 0.00000050, "input_cost_per_token": 0.00000015,
"output_cost_per_token": 0.0000015, "output_cost_per_token": 0.0000006,
"litellm_provider": "cohere_chat",
"mode": "chat",
"supports_function_calling": true
},
"command-r-08-2024": {
"max_tokens": 4096,
"max_input_tokens": 128000,
"max_output_tokens": 4096,
"input_cost_per_token": 0.00000015,
"output_cost_per_token": 0.0000006,
"litellm_provider": "cohere_chat", "litellm_provider": "cohere_chat",
"mode": "chat", "mode": "chat",
"supports_function_calling": true "supports_function_calling": true
@ -2684,8 +2693,18 @@
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 128000, "max_input_tokens": 128000,
"max_output_tokens": 4096, "max_output_tokens": 4096,
"input_cost_per_token": 0.000003, "input_cost_per_token": 0.0000025,
"output_cost_per_token": 0.000015, "output_cost_per_token": 0.00001,
"litellm_provider": "cohere_chat",
"mode": "chat",
"supports_function_calling": true
},
"command-r-plus-08-2024": {
"max_tokens": 4096,
"max_input_tokens": 128000,
"max_output_tokens": 4096,
"input_cost_per_token": 0.0000025,
"output_cost_per_token": 0.00001,
"litellm_provider": "cohere_chat", "litellm_provider": "cohere_chat",
"mode": "chat", "mode": "chat",
"supports_function_calling": true "supports_function_calling": true
@ -2699,25 +2718,7 @@
"litellm_provider": "cohere", "litellm_provider": "cohere",
"mode": "completion" "mode": "completion"
}, },
"command": { "command": {
"max_tokens": 4096,
"max_input_tokens": 4096,
"max_output_tokens": 4096,
"input_cost_per_token": 0.000015,
"output_cost_per_token": 0.000015,
"litellm_provider": "cohere",
"mode": "completion"
},
"command-medium-beta": {
"max_tokens": 4096,
"max_input_tokens": 4096,
"max_output_tokens": 4096,
"input_cost_per_token": 0.000015,
"output_cost_per_token": 0.000015,
"litellm_provider": "cohere",
"mode": "completion"
},
"command-xlarge-beta": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5235,7 +5236,7 @@
"supports_function_calling": true, "supports_function_calling": true,
"source": "https://fireworks.ai/pricing" "source": "https://fireworks.ai/pricing"
}, },
"anyscale/mistralai/Mistral-7B-Instruct-v0.1": { "anyscale/mistralai/Mistral-7B-Instruct-v0.1": {
"max_tokens": 16384, "max_tokens": 16384,
"max_input_tokens": 16384, "max_input_tokens": 16384,
"max_output_tokens": 16384, "max_output_tokens": 16384,
@ -5245,8 +5246,8 @@
"mode": "chat", "mode": "chat",
"supports_function_calling": true, "supports_function_calling": true,
"source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/mistralai-Mistral-7B-Instruct-v0.1" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/mistralai-Mistral-7B-Instruct-v0.1"
}, },
"anyscale/mistralai/Mixtral-8x7B-Instruct-v0.1": { "anyscale/mistralai/Mixtral-8x7B-Instruct-v0.1": {
"max_tokens": 16384, "max_tokens": 16384,
"max_input_tokens": 16384, "max_input_tokens": 16384,
"max_output_tokens": 16384, "max_output_tokens": 16384,
@ -5256,8 +5257,8 @@
"mode": "chat", "mode": "chat",
"supports_function_calling": true, "supports_function_calling": true,
"source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/mistralai-Mixtral-8x7B-Instruct-v0.1" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/mistralai-Mixtral-8x7B-Instruct-v0.1"
}, },
"anyscale/mistralai/Mixtral-8x22B-Instruct-v0.1": { "anyscale/mistralai/Mixtral-8x22B-Instruct-v0.1": {
"max_tokens": 65536, "max_tokens": 65536,
"max_input_tokens": 65536, "max_input_tokens": 65536,
"max_output_tokens": 65536, "max_output_tokens": 65536,
@ -5267,8 +5268,8 @@
"mode": "chat", "mode": "chat",
"supports_function_calling": true, "supports_function_calling": true,
"source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/mistralai-Mixtral-8x22B-Instruct-v0.1" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/mistralai-Mixtral-8x22B-Instruct-v0.1"
}, },
"anyscale/HuggingFaceH4/zephyr-7b-beta": { "anyscale/HuggingFaceH4/zephyr-7b-beta": {
"max_tokens": 16384, "max_tokens": 16384,
"max_input_tokens": 16384, "max_input_tokens": 16384,
"max_output_tokens": 16384, "max_output_tokens": 16384,
@ -5276,8 +5277,8 @@
"output_cost_per_token": 0.00000015, "output_cost_per_token": 0.00000015,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat" "mode": "chat"
}, },
"anyscale/google/gemma-7b-it": { "anyscale/google/gemma-7b-it": {
"max_tokens": 8192, "max_tokens": 8192,
"max_input_tokens": 8192, "max_input_tokens": 8192,
"max_output_tokens": 8192, "max_output_tokens": 8192,
@ -5286,8 +5287,8 @@
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat", "mode": "chat",
"source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/google-gemma-7b-it" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/google-gemma-7b-it"
}, },
"anyscale/meta-llama/Llama-2-7b-chat-hf": { "anyscale/meta-llama/Llama-2-7b-chat-hf": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5295,8 +5296,8 @@
"output_cost_per_token": 0.00000015, "output_cost_per_token": 0.00000015,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat" "mode": "chat"
}, },
"anyscale/meta-llama/Llama-2-13b-chat-hf": { "anyscale/meta-llama/Llama-2-13b-chat-hf": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5304,8 +5305,8 @@
"output_cost_per_token": 0.00000025, "output_cost_per_token": 0.00000025,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat" "mode": "chat"
}, },
"anyscale/meta-llama/Llama-2-70b-chat-hf": { "anyscale/meta-llama/Llama-2-70b-chat-hf": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5313,8 +5314,8 @@
"output_cost_per_token": 0.000001, "output_cost_per_token": 0.000001,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat" "mode": "chat"
}, },
"anyscale/codellama/CodeLlama-34b-Instruct-hf": { "anyscale/codellama/CodeLlama-34b-Instruct-hf": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5322,8 +5323,8 @@
"output_cost_per_token": 0.000001, "output_cost_per_token": 0.000001,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat" "mode": "chat"
}, },
"anyscale/codellama/CodeLlama-70b-Instruct-hf": { "anyscale/codellama/CodeLlama-70b-Instruct-hf": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5331,9 +5332,9 @@
"output_cost_per_token": 0.000001, "output_cost_per_token": 0.000001,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat", "mode": "chat",
"source" : "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/codellama-CodeLlama-70b-Instruct-hf" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/codellama-CodeLlama-70b-Instruct-hf"
}, },
"anyscale/meta-llama/Meta-Llama-3-8B-Instruct": { "anyscale/meta-llama/Meta-Llama-3-8B-Instruct": {
"max_tokens": 8192, "max_tokens": 8192,
"max_input_tokens": 8192, "max_input_tokens": 8192,
"max_output_tokens": 8192, "max_output_tokens": 8192,
@ -5342,8 +5343,8 @@
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat", "mode": "chat",
"source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/meta-llama-Meta-Llama-3-8B-Instruct" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/meta-llama-Meta-Llama-3-8B-Instruct"
}, },
"anyscale/meta-llama/Meta-Llama-3-70B-Instruct": { "anyscale/meta-llama/Meta-Llama-3-70B-Instruct": {
"max_tokens": 8192, "max_tokens": 8192,
"max_input_tokens": 8192, "max_input_tokens": 8192,
"max_output_tokens": 8192, "max_output_tokens": 8192,
@ -5351,9 +5352,9 @@
"output_cost_per_token": 0.00000100, "output_cost_per_token": 0.00000100,
"litellm_provider": "anyscale", "litellm_provider": "anyscale",
"mode": "chat", "mode": "chat",
"source" : "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/meta-llama-Meta-Llama-3-70B-Instruct" "source": "https://docs.anyscale.com/preview/endpoints/text-generation/supported-models/meta-llama-Meta-Llama-3-70B-Instruct"
}, },
"cloudflare/@cf/meta/llama-2-7b-chat-fp16": { "cloudflare/@cf/meta/llama-2-7b-chat-fp16": {
"max_tokens": 3072, "max_tokens": 3072,
"max_input_tokens": 3072, "max_input_tokens": 3072,
"max_output_tokens": 3072, "max_output_tokens": 3072,
@ -5361,8 +5362,8 @@
"output_cost_per_token": 0.000001923, "output_cost_per_token": 0.000001923,
"litellm_provider": "cloudflare", "litellm_provider": "cloudflare",
"mode": "chat" "mode": "chat"
}, },
"cloudflare/@cf/meta/llama-2-7b-chat-int8": { "cloudflare/@cf/meta/llama-2-7b-chat-int8": {
"max_tokens": 2048, "max_tokens": 2048,
"max_input_tokens": 2048, "max_input_tokens": 2048,
"max_output_tokens": 2048, "max_output_tokens": 2048,
@ -5370,8 +5371,8 @@
"output_cost_per_token": 0.000001923, "output_cost_per_token": 0.000001923,
"litellm_provider": "cloudflare", "litellm_provider": "cloudflare",
"mode": "chat" "mode": "chat"
}, },
"cloudflare/@cf/mistral/mistral-7b-instruct-v0.1": { "cloudflare/@cf/mistral/mistral-7b-instruct-v0.1": {
"max_tokens": 8192, "max_tokens": 8192,
"max_input_tokens": 8192, "max_input_tokens": 8192,
"max_output_tokens": 8192, "max_output_tokens": 8192,
@ -5379,8 +5380,8 @@
"output_cost_per_token": 0.000001923, "output_cost_per_token": 0.000001923,
"litellm_provider": "cloudflare", "litellm_provider": "cloudflare",
"mode": "chat" "mode": "chat"
}, },
"cloudflare/@hf/thebloke/codellama-7b-instruct-awq": { "cloudflare/@hf/thebloke/codellama-7b-instruct-awq": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"max_output_tokens": 4096, "max_output_tokens": 4096,
@ -5388,8 +5389,8 @@
"output_cost_per_token": 0.000001923, "output_cost_per_token": 0.000001923,
"litellm_provider": "cloudflare", "litellm_provider": "cloudflare",
"mode": "chat" "mode": "chat"
}, },
"voyage/voyage-01": { "voyage/voyage-01": {
"max_tokens": 4096, "max_tokens": 4096,
"max_input_tokens": 4096, "max_input_tokens": 4096,
"input_cost_per_token": 0.0000001, "input_cost_per_token": 0.0000001,
@ -5494,7 +5495,6 @@
"litellm_provider": "databricks", "litellm_provider": "databricks",
"mode": "chat", "mode": "chat",
"source": "https://www.databricks.com/product/pricing/foundation-model-serving" "source": "https://www.databricks.com/product/pricing/foundation-model-serving"
}, },
"databricks/databricks-mixtral-8x7b-instruct": { "databricks/databricks-mixtral-8x7b-instruct": {
"max_tokens": 4096, "max_tokens": 4096,