Skip to content

Open-source Python library for AI browser agents, plus Browser Use Cloud: stealth browsers and a hosted agent API.

model pricing

Gregor Žunič61b15ac243ff
diff --git a/browser_use/tokens/custom_pricing.py b/browser_use/tokens/custom_pricing.py
new file mode 100644
index 000000000..2be330a86
--- /dev/null
+++ b/browser_use/tokens/custom_pricing.py
@@ -0,0 +1,30 @@
+"""
+Custom model pricing for models not available in LiteLLM's pricing data.
+
+Prices are per token (not per 1M tokens).
+"""
+
+from typing import Any
+
+# Custom model pricing data
+# Format matches LiteLLM's model_prices_and_context_window.json structure
+CUSTOM_MODEL_PRICING: dict[str, dict[str, Any]] = {
+ 'browser-use/fast': {
+ 'input_cost_per_token': 0.50 / 1_000_000, # $0.50 per 1M tokens
+ 'output_cost_per_token': 3.00 / 1_000_000, # $3.00 per 1M tokens
+ 'cache_read_input_token_cost': 0.10 / 1_000_000, # $0.10 per 1M tokens
+ 'cache_creation_input_token_cost': None, # Not specified
+ 'max_tokens': None, # Not specified
+ 'max_input_tokens': None, # Not specified
+ 'max_output_tokens': None, # Not specified
+ },
+ 'browser-use/smart': {
+ 'input_cost_per_token': 0.50 / 1_000_000, # $0.50 per 1M tokens
+ 'output_cost_per_token': 3.00 / 1_000_000, # $3.00 per 1M tokens
+ 'cache_read_input_token_cost': 0.10 / 1_000_000, # $0.10 per 1M tokens
+ 'cache_creation_input_token_cost': None, # Not specified
+ 'max_tokens': None, # Not specified
+ 'max_input_tokens': None, # Not specified
+ 'max_output_tokens': None, # Not specified
+ },
+}
diff --git a/browser_use/tokens/mappings.py b/browser_use/tokens/mappings.py
new file mode 100644
index 000000000..01a5970b1
--- /dev/null
+++ b/browser_use/tokens/mappings.py
@@ -0,0 +1,4 @@
+# Mapping from model_name to LiteLLM model name
+MODEL_TO_LITELLM: dict[str, str] = {
+ 'gemini-flash-latest': 'gemini/gemini-flash-latest',
+}
diff --git a/browser_use/tokens/service.py b/browser_use/tokens/service.py
index 9b32e8294..b4277d2e4 100644
--- a/browser_use/tokens/service.py
+++ b/browser_use/tokens/service.py
@@ -18,6 +18,8 @@ from dotenv import load_dotenv
from browser_use.llm.base import BaseChatModel
from browser_use.llm.views import ChatInvokeUsage
+from browser_use.tokens.custom_pricing import CUSTOM_MODEL_PRICING
+from browser_use.tokens.mappings import MODEL_TO_LITELLM
from browser_use.tokens.views import (
CachedPricingData,
ModelPricing,
@@ -162,10 +164,27 @@ class TokenCost:
if not self._initialized:
await self.initialize()
- if not self._pricing_data or model_name not in self._pricing_data:
+ # Check custom pricing first
+ if model_name in CUSTOM_MODEL_PRICING:
+ data = CUSTOM_MODEL_PRICING[model_name]
+ return ModelPricing(
+ model=model_name,
+ input_cost_per_token=data.get('input_cost_per_token'),
+ output_cost_per_token=data.get('output_cost_per_token'),
+ max_tokens=data.get('max_tokens'),
+ max_input_tokens=data.get('max_input_tokens'),
+ max_output_tokens=data.get('max_output_tokens'),
+ cache_read_input_token_cost=data.get('cache_read_input_token_cost'),
+ cache_creation_input_token_cost=data.get('cache_creation_input_token_cost'),
+ )
+
+ # Map model name to LiteLLM model name if needed
+ litellm_model_name = MODEL_TO_LITELLM.get(model_name, model_name)
+
+ if not self._pricing_data or litellm_model_name not in self._pricing_data:
return None
- data = self._pricing_data[model_name]
+ data = self._pricing_data[litellm_model_name]
return ModelPricing(
model=model_name,
input_cost_per_token=data.get('input_cost_per_token'),
@@ -323,11 +342,11 @@ class TokenCost:
# Track usage if available (no await needed since add_usage is now sync)
if result.usage:
- usage = token_cost_service.add_usage(llm.model, result.usage)
+ usage = token_cost_service.add_usage(llm.name, result.usage)
logger.debug(f'Token cost service: {usage}')
- asyncio.create_task(token_cost_service._log_usage(llm.model, usage))
+ asyncio.create_task(token_cost_service._log_usage(llm.name, usage))
# else:
# await token_cost_service._log_non_usage_llm(llm)