Skip to content

🌟 The Multi-Agent Framework: First AI Software Company, Towards Natural Language Programming

update close-source latest model usage

better629f7dd8c965e40
diff --git a/metagpt/provider/openai_api.py b/metagpt/provider/openai_api.py
index 31907d9e8..85d5ef50e 100644
--- a/metagpt/provider/openai_api.py
+++ b/metagpt/provider/openai_api.py
@@ -37,7 +37,6 @@ from metagpt.utils.token_counter import (
count_input_tokens,
count_output_tokens,
get_max_completion_tokens,
- get_openrouter_tokens,
)
@@ -92,6 +91,7 @@ class OpenAILLM(BaseLLM):
)
usage = None
collected_messages = []
+ has_finished = False
async for chunk in response:
chunk_message = chunk.choices[0].delta.content or "" if chunk.choices else "" # extract the message
finish_reason = (
@@ -99,6 +99,10 @@ class OpenAILLM(BaseLLM):
)
log_llm_stream(chunk_message)
collected_messages.append(chunk_message)
+ if has_finished:
+ # for oneapi, there has a usage chunk after finish_reason not none chunk
+ if hasattr(chunk, "usage"):
+ usage = CompletionUsage(**chunk.usage)
if finish_reason:
if hasattr(chunk, "usage") and chunk.usage is not None:
# Some services have usage as an attribute of the chunk, such as Fireworks
@@ -109,9 +113,10 @@ class OpenAILLM(BaseLLM):
elif hasattr(chunk.choices[0], "usage"):
# The usage of some services is an attribute of chunk.choices[0], such as Moonshot
usage = CompletionUsage(**chunk.choices[0].usage)
- elif "openrouter.ai" in self.config.base_url:
+ elif "openrouter.ai" in self.config.base_url and hasattr(chunk, "usage") and chunk.usage is not None:
# due to it get token cost from api
- usage = await get_openrouter_tokens(chunk)
+ usage = chunk.usage
+ has_finished = True
log_llm_stream("\n")
full_reply_content = "".join(collected_messages)
diff --git a/metagpt/utils/token_counter.py b/metagpt/utils/token_counter.py
index 49d62b69a..90fd9e960 100644
--- a/metagpt/utils/token_counter.py
+++ b/metagpt/utils/token_counter.py
@@ -41,11 +41,19 @@ TOKEN_COSTS = {
"gpt-4-1106-vision-preview": {"prompt": 0.01, "completion": 0.03},
"gpt-4o": {"prompt": 0.005, "completion": 0.015},
"gpt-4o-mini": {"prompt": 0.00015, "completion": 0.0006},
+ "gpt-4o-mini-2024-07-18": {"prompt": 0.00015, "completion": 0.0006},
"gpt-4o-2024-05-13": {"prompt": 0.005, "completion": 0.015},
+ "gpt-4o-2024-08-06": {"prompt": 0.0025, "completion": 0.01},
+ "o1-preview": {"prompt": 0.015, "completion": 0.06},
+ "o1-preview-2024-09-12": {"prompt": 0.015, "completion": 0.06},
+ "o1-mini": {"prompt": 0.003, "completion": 0.012},
+ "o1-mini-2024-09-12": {"prompt": 0.003, "completion": 0.012},
"text-embedding-ada-002": {"prompt": 0.0004, "completion": 0.0},
"glm-3-turbo": {"prompt": 0.0007, "completion": 0.0007}, # 128k version, prompt + completion tokens=0.005ï¿¥/k-tokens
"glm-4": {"prompt": 0.014, "completion": 0.014}, # 128k version, prompt + completion tokens=0.1ï¿¥/k-tokens
- "gemini-pro": {"prompt": 0.00025, "completion": 0.0005},
+ "gemini-1.5-flash": {"prompt": 0.000075, "completion": 0.0003},
+ "gemini-1.5-pro": {"prompt": 0.0035, "completion": 0.0105},
+ "gemini-1.0-pro": {"prompt": 0.0005, "completion": 0.0015},
"moonshot-v1-8k": {"prompt": 0.012, "completion": 0.012}, # prompt + completion tokens=0.012ï¿¥/k-tokens
"moonshot-v1-32k": {"prompt": 0.024, "completion": 0.024},
"moonshot-v1-128k": {"prompt": 0.06, "completion": 0.06},
@@ -69,15 +77,20 @@ TOKEN_COSTS = {
"llama3-70b-8192": {"prompt": 0.0059, "completion": 0.0079},
"openai/gpt-3.5-turbo-0125": {"prompt": 0.0005, "completion": 0.0015},
"openai/gpt-4-turbo-preview": {"prompt": 0.01, "completion": 0.03},
+ "openai/o1-preview": {"prompt": 0.015, "completion": 0.06},
+ "openai/o1-mini": {"prompt": 0.003, "completion": 0.012},
+ "anthropic/claude-3-opus": {"prompt": 0.015, "completion": 0.075},
+ "anthropic/claude-3.5-sonnet": {"prompt": 0.003, "completion": 0.015},
+ "google/gemini-pro-1.5": {"prompt": 0.0025, "completion": 0.0075}, # for openrouter, end
"deepseek-chat": {"prompt": 0.00014, "completion": 0.00028},
"deepseek-coder": {"prompt": 0.00014, "completion": 0.00028},
# For ark model https://www.volcengine.com/docs/82379/1099320
- "doubao-lite-4k-240515": {"prompt": 0.000042, "completion": 0.000084},
- "doubao-lite-32k-240515": {"prompt": 0.000042, "completion": 0.000084},
- "doubao-lite-128k-240515": {"prompt": 0.00011, "completion": 0.00013},
- "doubao-pro-4k-240515": {"prompt": 0.00011, "completion": 0.00028},
- "doubao-pro-32k-240515": {"prompt": 0.00011, "completion": 0.00028},
- "doubao-pro-128k-240515": {"prompt": 0.0007, "completion": 0.0012},
+ "doubao-lite-4k-240515": {"prompt": 0.000043, "completion": 0.000086},
+ "doubao-lite-32k-240515": {"prompt": 0.000043, "completion": 0.000086},
+ "doubao-lite-128k-240515": {"prompt": 0.00011, "completion": 0.00014},
+ "doubao-pro-4k-240515": {"prompt": 0.00011, "completion": 0.00029},
+ "doubao-pro-32k-240515": {"prompt": 0.00011, "completion": 0.00029},
+ "doubao-pro-128k-240515": {"prompt": 0.0007, "completion": 0.0013},
"llama3-70b-llama3-70b-instruct": {"prompt": 0.0, "completion": 0.0},
"llama3-8b-llama3-8b-instruct": {"prompt": 0.0, "completion": 0.0},
}
@@ -142,10 +155,10 @@ Some new model published by Alibaba will be prioritized to be released on the Mo
Token price on Model Studio shows on https://help.aliyun.com/zh/model-studio/getting-started/models#ced16cb6cdfsy
"""
DASHSCOPE_TOKEN_COSTS = {
- "qwen2.5-72b-instruct": {"prompt": 0.004, "completion": 0.012}, # per 1k tokens
- "qwen2.5-32b-instruct": {"prompt": 0.0035, "completion": 0.007},
- "qwen2.5-14b-instruct": {"prompt": 0.002, "completion": 0.006},
- "qwen2.5-7b-instruct": {"prompt": 0.001, "completion": 0.002},
+ "qwen2.5-72b-instruct": {"prompt": 0.00057, "completion": 0.0017}, # per 1k tokens
+ "qwen2.5-32b-instruct": {"prompt": 0.0005, "completion": 0.001},
+ "qwen2.5-14b-instruct": {"prompt": 0.00029, "completion": 0.00086},
+ "qwen2.5-7b-instruct": {"prompt": 0.00014, "completion": 0.00029},
"qwen2.5-3b-instruct": {"prompt": 0.0, "completion": 0.0},
"qwen2.5-1.5b-instruct": {"prompt": 0.0, "completion": 0.0},
"qwen2.5-0.5b-instruct": {"prompt": 0.0, "completion": 0.0},
@@ -199,16 +212,24 @@ FIREWORKS_GRADE_TOKEN_COSTS = {
# https://console.volcengine.com/ark/region:ark+cn-beijing/model
DOUBAO_TOKEN_COSTS = {
- "doubao-lite": {"prompt": 0.0003, "completion": 0.0006},
- "doubao-lite-128k": {"prompt": 0.0008, "completion": 0.0010},
- "doubao-pro": {"prompt": 0.0008, "completion": 0.0020},
- "doubao-pro-128k": {"prompt": 0.0050, "completion": 0.0090},
+ "doubao-lite": {"prompt": 0.000043, "completion": 0.000086},
+ "doubao-lite-128k": {"prompt": 0.00011, "completion": 0.00014},
+ "doubao-pro": {"prompt": 0.00011, "completion": 0.00029},
+ "doubao-pro-128k": {"prompt": 0.00071, "completion": 0.0013},
+ "doubao-pro-256k": {"prompt": 0.00071, "completion": 0.0013},
}
# https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo
TOKEN_MAX = {
- "gpt-4o-2024-05-13": 128000,
+ "o1-preview": 128000,
+ "o1-preview-2024-09-12": 128000,
+ "o1-mini": 128000,
+ "o1-mini-2024-09-12": 128000,
"gpt-4o": 128000,
+ "gpt-4o-2024-05-13": 128000,
+ "gpt-4o-2024-08-06": 128000,
+ "gpt-4o-mini-2024-07-18": 128000,
+ "gpt-4o-mini": 128000,
"gpt-4-turbo-2024-04-09": 128000,
"gpt-4-0125-preview": 128000,
"gpt-4-turbo-preview": 128000,
@@ -220,7 +241,6 @@ TOKEN_MAX = {
"gpt-4-0613": 8192,
"gpt-4-32k": 32768,
"gpt-4-32k-0613": 32768,
- "gpt-4o-mini": 128000,
"gpt-3.5-turbo-0125": 16385,
"gpt-3.5-turbo": 16385,
"gpt-3.5-turbo-1106": 16385,
@@ -231,7 +251,9 @@ TOKEN_MAX = {
"text-embedding-ada-002": 8192,
"glm-3-turbo": 128000,
"glm-4": 128000,
- "gemini-pro": 32768,
+ "gemini-1.5-flash": 1000000,
+ "gemini-1.5-pro": 2000000,
+ "gemini-1.0-pro": 32000,
"moonshot-v1-8k": 8192,
"moonshot-v1-32k": 32768,
"moonshot-v1-128k": 128000,
@@ -255,6 +277,11 @@ TOKEN_MAX = {
"llama3-70b-8192": 8192,
"openai/gpt-3.5-turbo-0125": 16385,
"openai/gpt-4-turbo-preview": 128000,
+ "openai/o1-preview": 128000,
+ "openai/o1-mini": 128000,
+ "anthropic/claude-3-opus": 200000,
+ "anthropic/claude-3.5-sonnet": 200000,
+ "google/gemini-pro-1.5": 4000000,
"deepseek-chat": 32768,
"deepseek-coder": 16385,
"doubao-lite-4k-240515": 4000,
@@ -375,7 +402,13 @@ def count_input_tokens(messages, model="gpt-3.5-turbo-0125"):
"gpt-4-1106-vision-preview",
"gpt-4o",
"gpt-4o-2024-05-13",
+ "gpt-4o-2024-08-06",
"gpt-4o-mini",
+ "gpt-4o-mini-2024-07-18",
+ "o1-preview",
+ "o1-preview-2024-09-12",
+ "o1-mini",
+ "o1-mini-2024-09-12",
}:
tokens_per_message = 3 # # every reply is primed with <|start|>assistant<|message|>
tokens_per_name = 1