scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-07 08:26:10 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 36

fix(model_prices): correct Claude 3 1-hour cache write rates, add W&B and Vertex retirement dates

Browse source 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
Devin AI 2026-08-24 13:13:19 +00:00
parent 64eafff28f
commit c5da8eb41c
2 changed files with 14 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
litellm/model_prices_and_context_window_backup.json
View file

@ -12306,7 +12306,7 @@
},
"claude-3-haiku-20240307": {
"cache_creation_input_token_cost": 3e-07,
"cache_creation_input_token_cost_above_1hr": 6e-06,
"cache_creation_input_token_cost_above_1hr": 5e-07,
"cache_read_input_token_cost": 3e-08,
"deprecation_date": "2026-04-20",
"input_cost_per_token": 2.5e-07,
@ -12325,7 +12325,7 @@
},
"claude-3-opus-20240229": {
"cache_creation_input_token_cost": 1.875e-05,
"cache_creation_input_token_cost_above_1hr": 6e-06,
"cache_creation_input_token_cost_above_1hr": 3e-05,
"cache_read_input_token_cost": 1.5e-06,
"deprecation_date": "2026-01-05",
"input_cost_per_token": 1.5e-05,
@ -31641,6 +31641,7 @@
"supports_vision": false
},
"multimodalembedding": {
"deprecation_date": "2027-04-01",
"input_cost_per_character": 2e-07,
"input_cost_per_image": 0.0001,
"input_cost_per_token": 8e-07,
@ -42448,6 +42449,7 @@
"source": "https://wandb.ai/site/pricing/tokens/"
},
"wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct": {
"deprecation_date": "2026-08-25",
"max_tokens": 262000,
"max_input_tokens": 262000,
"max_output_tokens": 262000,
@ -42458,6 +42460,7 @@
"source": "https://wandb.ai/site/pricing/tokens/"
},
"wandb/moonshotai/Kimi-K2-Instruct": {
"deprecation_date": "2026-03-04",
"max_tokens": 128000,
"max_input_tokens": 128000,
"max_output_tokens": 128000,
@ -42482,6 +42485,7 @@
"supports_vision": true
},
"wandb/MiniMaxAI/MiniMax-M2.5": {
"deprecation_date": "2026-08-25",
"max_tokens": 197000,
"max_input_tokens": 197000,
"max_output_tokens": 197000,
@ -53247,6 +53251,7 @@
"supports_vision": true
},
"wandb/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": {
"deprecation_date": "2026-08-25",
"input_cost_per_token": 2e-07,
"output_cost_per_token": 8e-07,
"max_input_tokens": 262000,

9
model_prices_and_context_window.json
View file

@ -12306,7 +12306,7 @@
},
"claude-3-haiku-20240307": {
"cache_creation_input_token_cost": 3e-07,
"cache_creation_input_token_cost_above_1hr": 6e-06,
"cache_creation_input_token_cost_above_1hr": 5e-07,
"cache_read_input_token_cost": 3e-08,
"deprecation_date": "2026-04-20",
"input_cost_per_token": 2.5e-07,
@ -12325,7 +12325,7 @@
},
"claude-3-opus-20240229": {
"cache_creation_input_token_cost": 1.875e-05,
"cache_creation_input_token_cost_above_1hr": 6e-06,
"cache_creation_input_token_cost_above_1hr": 3e-05,
"cache_read_input_token_cost": 1.5e-06,
"deprecation_date": "2026-01-05",
"input_cost_per_token": 1.5e-05,
@ -31641,6 +31641,7 @@
"supports_vision": false
},
"multimodalembedding": {
"deprecation_date": "2027-04-01",
"input_cost_per_character": 2e-07,
"input_cost_per_image": 0.0001,
"input_cost_per_token": 8e-07,
@ -42448,6 +42449,7 @@
"source": "https://wandb.ai/site/pricing/tokens/"
},
"wandb/Qwen/Qwen3-Coder-480B-A35B-Instruct": {
"deprecation_date": "2026-08-25",
"max_tokens": 262000,
"max_input_tokens": 262000,
"max_output_tokens": 262000,
@ -42458,6 +42460,7 @@
"source": "https://wandb.ai/site/pricing/tokens/"
},
"wandb/moonshotai/Kimi-K2-Instruct": {
"deprecation_date": "2026-03-04",
"max_tokens": 128000,
"max_input_tokens": 128000,
"max_output_tokens": 128000,
@ -42482,6 +42485,7 @@
"supports_vision": true
},
"wandb/MiniMaxAI/MiniMax-M2.5": {
"deprecation_date": "2026-08-25",
"max_tokens": 197000,
"max_input_tokens": 197000,
"max_output_tokens": 197000,
@ -53247,6 +53251,7 @@
"supports_vision": true
},
"wandb/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": {
"deprecation_date": "2026-08-25",
"input_cost_per_token": 2e-07,
"output_cost_per_token": 8e-07,
"max_input_tokens": 262000,