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

feat(pricing): add gemini-omni-flash-preview with video output token pricing

Browse source
This commit is contained in:
mateo-berri 2026-07-14 14:04:51 -07:00
parent edd3bce0ec
commit 598fa9d64d
8 changed files with 276 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

35
litellm/litellm_core_utils/llm_cost_calc/utils.py
View file

@ -445,6 +445,7 @@ class PromptTokensDetailsResult(TypedDict):
text_tokens: int
audio_tokens: int
image_tokens: int
video_tokens: int
character_count: int
image_count: int
video_length_seconds: float
@ -473,6 +474,7 @@ def _parse_prompt_tokens_details(usage: Usage) -> PromptTokensDetailsResult:
)
audio_tokens = cast(Optional[int], getattr(usage.prompt_tokens_details, "audio_tokens", 0)) or 0
image_tokens = cast(Optional[int], getattr(usage.prompt_tokens_details, "image_tokens", 0)) or 0
video_tokens = _coerce_token_count(getattr(usage.prompt_tokens_details, "video_tokens", 0))
character_count = (
cast(
Optional[int],
@ -503,6 +505,7 @@ def _parse_prompt_tokens_details(usage: Usage) -> PromptTokensDetailsResult:
text_tokens=text_tokens,
audio_tokens=audio_tokens,
image_tokens=image_tokens,
video_tokens=video_tokens,
character_count=character_count,
image_count=image_count,
video_length_seconds=float(video_length_seconds),
@ -515,6 +518,7 @@ class CompletionTokensDetailsResult(TypedDict):
text_tokens: int
reasoning_tokens: int
image_tokens: int
video_tokens: int
def _parse_completion_tokens_details(usage: Usage) -> CompletionTokensDetailsResult:
@ -546,12 +550,14 @@ def _parse_completion_tokens_details(usage: Usage) -> CompletionTokensDetailsRes
)
or 0
)
video_tokens = _coerce_token_count(getattr(usage.completion_tokens_details, "video_tokens", 0))
return CompletionTokensDetailsResult(
audio_tokens=audio_tokens,
text_tokens=text_tokens,
reasoning_tokens=reasoning_tokens,
image_tokens=image_tokens,
video_tokens=video_tokens,
)
@ -586,6 +592,13 @@ def _calculate_input_cost(
image_token_cost_key = "input_cost_per_token"
prompt_cost += calculate_cost_component(model_info, image_token_cost_key, prompt_tokens_details["image_tokens"])
### VIDEO TOKEN COST
if prompt_tokens_details["video_tokens"]:
video_token_cost_key = "input_cost_per_video_token"
if model_info.get(video_token_cost_key) is None:
video_token_cost_key = "input_cost_per_token"
prompt_cost += calculate_cost_component(model_info, video_token_cost_key, prompt_tokens_details["video_tokens"])
### CACHE WRITING COST - Now uses tiered pricing
if (
prompt_tokens_details["cache_creation_tokens"]
@ -698,6 +711,7 @@ def generic_cost_per_token(
text_tokens=usage.prompt_tokens,
audio_tokens=0,
image_tokens=0,
video_tokens=0,
character_count=0,
image_count=0,
video_length_seconds=0.0,
@ -716,13 +730,14 @@ def generic_cost_per_token(
audio_tokens = prompt_tokens_details["audio_tokens"]
cache_creation = prompt_tokens_details["cache_creation_tokens"]
image_tokens = prompt_tokens_details["image_tokens"]
video_tokens = prompt_tokens_details["video_tokens"]
# Check for double-counting: sum of details > prompt_tokens means overlap
total_details = text_tokens + cache_hit + audio_tokens + cache_creation + image_tokens
total_details = text_tokens + cache_hit + audio_tokens + cache_creation + image_tokens + video_tokens
has_double_counting = cache_hit > 0 and total_details > usage.prompt_tokens
if (text_tokens == 0 and prompt_tokens_details["image_count"] == 0) or has_double_counting:
text_tokens = usage.prompt_tokens - cache_hit - audio_tokens - cache_creation - image_tokens
text_tokens = usage.prompt_tokens - cache_hit - audio_tokens - cache_creation - image_tokens - video_tokens
# Clamp to zero: inconsistent streaming usage
if text_tokens < 0:
text_tokens = 0
@ -751,6 +766,7 @@ def generic_cost_per_token(
audio_tokens = 0
reasoning_tokens = 0
image_tokens = 0
video_tokens = 0
is_text_tokens_total = False
if usage.completion_tokens_details is not None:
completion_tokens_details = _parse_completion_tokens_details(usage)
@ -758,19 +774,20 @@ def generic_cost_per_token(
text_tokens = completion_tokens_details["text_tokens"]
reasoning_tokens = completion_tokens_details["reasoning_tokens"]
image_tokens = completion_tokens_details["image_tokens"]
video_tokens = completion_tokens_details["video_tokens"]
# Handle text_tokens calculation:
# 1. If text_tokens is explicitly provided and > 0, use it
# 2. If there's a breakdown (reasoning/audio/image tokens), calculate text_tokens as the remainder
# 2. If there's a breakdown (reasoning/audio/image/video tokens), calculate text_tokens as the remainder
# 3. If no breakdown at all, assume all completion_tokens are text_tokens
has_token_breakdown = image_tokens > 0 or audio_tokens > 0 or reasoning_tokens > 0
has_token_breakdown = image_tokens > 0 or audio_tokens > 0 or reasoning_tokens > 0 or video_tokens > 0
if text_tokens == 0:
if has_token_breakdown:
# Calculate text tokens as remainder when we have a breakdown
# This handles cases like OpenAI's reasoning models where text_tokens isn't provided
text_tokens = max(
0,
usage.completion_tokens - reasoning_tokens - audio_tokens - image_tokens,
usage.completion_tokens - reasoning_tokens - audio_tokens - image_tokens - video_tokens,
)
else:
# No breakdown at all, all tokens are text tokens
@ -803,6 +820,14 @@ def generic_cost_per_token(
)
completion_cost += float(image_tokens) * _output_cost_per_image_token
## VIDEO COST
if not is_text_tokens_total and video_tokens and video_tokens > 0:
_output_cost_per_video_token = _get_cost_per_unit(model_info, "output_cost_per_video_token", None)
_output_cost_per_video_token = (
_output_cost_per_video_token if _output_cost_per_video_token is not None else completion_base_cost
)
completion_cost += float(video_tokens) * _output_cost_per_video_token
## REGIONAL DATA-RESIDENCY UPLIFT
# Applied as a flat multiplier across all token costs for the request
# when the upstream is a regionalized OpenAI host (eu./us.api.openai.com).

64
litellm/model_prices_and_context_window_backup.json
View file

@ -19600,6 +19600,39 @@
},
"web_search_billing_unit": "per_query"
},
"gemini/gemini-omni-flash-preview": {
"input_cost_per_audio_token": 1.5e-06,
"input_cost_per_token": 1.5e-06,
"litellm_provider": "gemini",
"max_input_tokens": 1048576,
"max_output_tokens": 65535,
"max_tokens": 65535,
"mode": "chat",
"output_cost_per_reasoning_token": 9e-06,
"output_cost_per_token": 9e-06,
"output_cost_per_video_token": 1.75e-05,
"rpm": 2000,
"source": "https://ai.google.dev/gemini-api/docs/pricing",
"supported_endpoints": [
"/v1/chat/completions"
],
"supported_modalities": [
"text",
"image",
"audio",
"video"
],
"supported_output_modalities": [
"text",
"video"
],
"supports_audio_input": true,
"supports_reasoning": true,
"supports_system_messages": true,
"supports_video_input": true,
"supports_vision": true,
"tpm": 800000
},
"gemini/gemini-3.1-pro-preview": {
"cache_read_input_token_cost": 2e-07,
"cache_read_input_token_cost_above_200k_tokens": 4e-07,
@ -19764,6 +19797,37 @@
},
"web_search_billing_unit": "per_query"
},
"gemini-omni-flash-preview": {
"input_cost_per_audio_token": 1.5e-06,
"input_cost_per_token": 1.5e-06,
"litellm_provider": "vertex_ai-language-models",
"max_input_tokens": 1048576,
"max_output_tokens": 65535,
"max_tokens": 65535,
"mode": "chat",
"output_cost_per_reasoning_token": 9e-06,
"output_cost_per_token": 9e-06,
"output_cost_per_video_token": 1.75e-05,
"source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/omni-flash-preview",
"supported_endpoints": [
"/v1/chat/completions"
],
"supported_modalities": [
"text",
"image",
"audio",
"video"
],
"supported_output_modalities": [
"text",
"video"
],
"supports_audio_input": true,
"supports_reasoning": true,
"supports_system_messages": true,
"supports_video_input": true,
"supports_vision": true
},
"gemini-3.5-flash": {
"cache_read_input_token_cost": 1.5e-07,
"input_cost_per_audio_token": 1e-06,

4
litellm/types/utils.py
View file

@ -209,6 +209,7 @@ class ModelInfoBase(ProviderSpecificModelInfo, total=False):
input_cost_per_query: Optional[float] # only for rerank models
input_cost_per_image: Optional[float] # only for vertex ai models
input_cost_per_image_token: Optional[float] # for gpt-image-1 and similar models
input_cost_per_video_token: Optional[float] # for gemini omni models with video input
input_cost_per_audio_per_second: Optional[float] # only for vertex ai models
input_cost_per_video_per_second: Optional[float] # only for vertex ai models
input_cost_per_second: Optional[float] # for OpenAI Speech models
@ -234,6 +235,7 @@ class ModelInfoBase(ProviderSpecificModelInfo, total=False):
output_cost_per_character_above_128k_tokens: Optional[float] # only for vertex ai models
output_cost_per_image: Optional[float]
output_cost_per_image_token: Optional[float]
output_cost_per_video_token: Optional[float] # for gemini omni models with video output
output_vector_size: Optional[int]
output_cost_per_reasoning_token: Optional[float]
output_cost_per_video_per_second: Optional[float] # only for vertex ai models
@ -3046,6 +3048,7 @@ class CustomPricingLiteLLMParams(BaseModel):
output_cost_per_character_above_128k_tokens: Optional[float] = None
output_cost_per_image: Optional[float] = None
output_cost_per_image_token: Optional[float] = None
output_cost_per_video_token: Optional[float] = None
output_cost_per_reasoning_token: Optional[float] = None
output_cost_per_video_per_second: Optional[float] = None
output_cost_per_audio_per_second: Optional[float] = None
@ -3055,6 +3058,7 @@ class CustomPricingLiteLLMParams(BaseModel):
cache_read_input_token_cost_above_272k_tokens: Optional[float] = None
cache_read_input_token_cost_above_512k_tokens: Optional[float] = None
input_cost_per_image_token: Optional[float] = None
input_cost_per_video_token: Optional[float] = None
input_cost_per_token_above_272k_tokens: Optional[float] = None
input_cost_per_token_above_512k_tokens: Optional[float] = None
output_cost_per_token_above_272k_tokens: Optional[float] = None

2
litellm/utils.py
View file

@ -5437,6 +5437,7 @@ def _get_model_info_helper(
input_cost_per_second=_model_info.get("input_cost_per_second", None),
input_cost_per_audio_token=_model_info.get("input_cost_per_audio_token", None),
input_cost_per_image_token=_model_info.get("input_cost_per_image_token", None),
input_cost_per_video_token=_model_info.get("input_cost_per_video_token", None),
input_cost_per_image=_model_info.get("input_cost_per_image", None),
input_cost_per_audio_per_second=_model_info.get("input_cost_per_audio_per_second", None),
input_cost_per_video_per_second=_model_info.get("input_cost_per_video_per_second", None),
@ -5480,6 +5481,7 @@ def _get_model_info_helper(
output_cost_per_video_per_second=_model_info.get("output_cost_per_video_per_second", None),
output_cost_per_image=_model_info.get("output_cost_per_image", None),
output_cost_per_image_token=_model_info.get("output_cost_per_image_token", None),
output_cost_per_video_token=_model_info.get("output_cost_per_video_token", None),
output_vector_size=_model_info.get("output_vector_size", None),
citation_cost_per_token=_model_info.get("citation_cost_per_token", None),
tiered_pricing=_model_info.get("tiered_pricing", None),

64
model_prices_and_context_window.json
View file

@ -19678,6 +19678,39 @@
},
"web_search_billing_unit": "per_query"
},
"gemini/gemini-omni-flash-preview": {
"input_cost_per_audio_token": 1.5e-06,
"input_cost_per_token": 1.5e-06,
"litellm_provider": "gemini",
"max_input_tokens": 1048576,
"max_output_tokens": 65535,
"max_tokens": 65535,
"mode": "chat",
"output_cost_per_reasoning_token": 9e-06,
"output_cost_per_token": 9e-06,
"output_cost_per_video_token": 1.75e-05,
"rpm": 2000,
"source": "https://ai.google.dev/gemini-api/docs/pricing",
"supported_endpoints": [
"/v1/chat/completions"
],
"supported_modalities": [
"text",
"image",
"audio",
"video"
],
"supported_output_modalities": [
"text",
"video"
],
"supports_audio_input": true,
"supports_reasoning": true,
"supports_system_messages": true,
"supports_video_input": true,
"supports_vision": true,
"tpm": 800000
},
"gemini/gemini-3.1-pro-preview": {
"cache_read_input_token_cost": 2e-07,
"cache_read_input_token_cost_above_200k_tokens": 4e-07,
@ -19842,6 +19875,37 @@
},
"web_search_billing_unit": "per_query"
},
"gemini-omni-flash-preview": {
"input_cost_per_audio_token": 1.5e-06,
"input_cost_per_token": 1.5e-06,
"litellm_provider": "vertex_ai-language-models",
"max_input_tokens": 1048576,
"max_output_tokens": 65535,
"max_tokens": 65535,
"mode": "chat",
"output_cost_per_reasoning_token": 9e-06,
"output_cost_per_token": 9e-06,
"output_cost_per_video_token": 1.75e-05,
"source": "https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/gemini/omni-flash-preview",
"supported_endpoints": [
"/v1/chat/completions"
],
"supported_modalities": [
"text",
"image",
"audio",
"video"
],
"supported_output_modalities": [
"text",
"video"
],
"supports_audio_input": true,
"supports_reasoning": true,
"supports_system_messages": true,
"supports_video_input": true,
"supports_vision": true
},
"gemini-3.5-flash": {
"cache_read_input_token_cost": 1.5e-07,
"input_cost_per_audio_token": 1e-06,

100
tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
View file

@ -270,6 +270,105 @@ def test_image_tokens_fallback_to_base_cost():
assert round(completion_cost, 12) == round(expected_completion_cost, 12)
def test_video_output_tokens_gemini_omni_flash_preview():
"""Video output tokens are billed at output_cost_per_video_token, not the text rate and not zero."""
model = "gemini-omni-flash-preview"
os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
litellm.model_cost = litellm.get_model_cost_map(url="")
text_tokens = 100
video_tokens = 46336
usage = Usage(
completion_tokens=text_tokens + video_tokens,
prompt_tokens=20,
total_tokens=20 + text_tokens + video_tokens,
completion_tokens_details=CompletionTokensDetailsWrapper(
text_tokens=text_tokens,
video_tokens=video_tokens,
),
prompt_tokens_details=PromptTokensDetailsWrapper(text_tokens=20),
)
model_cost_map = litellm.model_cost[f"gemini/{model}"]
assert model_cost_map["input_cost_per_token"] == 1.5e-06
assert model_cost_map["output_cost_per_token"] == 9e-06
assert model_cost_map["output_cost_per_video_token"] == 1.75e-05
prompt_cost, completion_cost = generic_cost_per_token(
model=model,
usage=usage,
custom_llm_provider="gemini",
)
assert round(prompt_cost, 10) == round(
model_cost_map["input_cost_per_token"] * usage.prompt_tokens,
10,
)
assert round(completion_cost, 10) == round(
(model_cost_map["output_cost_per_token"] * text_tokens)
+ (model_cost_map["output_cost_per_video_token"] * video_tokens),
10,
)
def test_video_input_tokens_gemini_omni_flash_preview():
"""Video input tokens are billed at the standard input rate instead of being dropped."""
model = "gemini-omni-flash-preview"
os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
litellm.model_cost = litellm.get_model_cost_map(url="")
usage = Usage(
completion_tokens=10,
prompt_tokens=10050,
total_tokens=10060,
completion_tokens_details=CompletionTokensDetailsWrapper(text_tokens=10),
prompt_tokens_details=PromptTokensDetailsWrapper(text_tokens=50, video_tokens=10000),
)
model_cost_map = litellm.model_cost[f"gemini/{model}"]
prompt_cost, _ = generic_cost_per_token(
model=model,
usage=usage,
custom_llm_provider="gemini",
)
assert round(prompt_cost, 10) == round(
model_cost_map["input_cost_per_token"] * usage.prompt_tokens,
10,
)
def test_video_tokens_fallback_to_base_cost():
"""Video output tokens fall back to the base output rate when output_cost_per_video_token is not set."""
from unittest.mock import patch
mock_model_info = {
"input_cost_per_token": 1e-6,
"output_cost_per_token": 2e-6,
}
usage = Usage(
completion_tokens=1720,
prompt_tokens=14,
total_tokens=1734,
completion_tokens_details=CompletionTokensDetailsWrapper(
text_tokens=600,
video_tokens=1120,
),
prompt_tokens_details=PromptTokensDetailsWrapper(text_tokens=14),
)
with patch(
"litellm.litellm_core_utils.llm_cost_calc.utils.get_model_info",
return_value=mock_model_info,
):
prompt_cost, completion_cost = generic_cost_per_token(
model="test-model", usage=usage, custom_llm_provider="gemini"
)
assert round(prompt_cost, 12) == round(14 * 1e-6, 12)
assert round(completion_cost, 12) == round((600 + 1120) * 2e-6, 12)
def test_generic_cost_per_token_above_200k_tokens():
# gemini-2.5-pro-exp-03-25 was removed; gemini-2.5-pro has same above-200k pricing
model = "gemini-2.5-pro"
@ -1086,6 +1185,7 @@ def test_cache_writing_cost_with_zero_creation_tokens_and_ephemeral_details():
"text_tokens": 0,
"audio_tokens": 0,
"image_tokens": 0,
"video_tokens": 0,
"character_count": 0,
"image_count": 0,
"video_length_seconds": 0.0,

4
tests/test_litellm/test_utils.py
View file

@ -615,6 +615,8 @@ def validate_model_cost_values(model_data, exceptions=None):
"input_cost_per_audio_token",
"output_cost_per_audio_token",
"output_cost_per_image_token",
"input_cost_per_video_token",
"output_cost_per_video_token",
"input_cost_per_audio_per_second",
"input_cost_per_video_per_second",
"input_cost_per_token_above_128k_tokens",
@ -732,6 +734,7 @@ def test_aaamodel_prices_and_context_window_json_is_valid():
"input_cost_per_image": {"type": "number"},
"input_cost_per_image_above_128k_tokens": {"type": "number"},
"input_cost_per_image_token": {"type": "number"},
"input_cost_per_video_token": {"type": "number"},
"input_cost_per_token_above_200k_tokens": {"type": "number"},
"input_cost_per_token_above_256k_tokens": {"type": "number"},
"input_cost_per_token_above_272k_tokens": {"type": "number"},
@ -807,6 +810,7 @@ def test_aaamodel_prices_and_context_window_json_is_valid():
"output_cost_per_character_above_128k_tokens": {"type": "number"},
"output_cost_per_image": {"type": "number"},
"output_cost_per_image_token": {"type": "number"},
"output_cost_per_video_token": {"type": "number"},
"output_cost_per_pixel": {"type": "number"},
"output_cost_per_second": {"type": "number"},
"output_cost_per_second_1080p": {"type": "number"},

8
ui/litellm-dashboard/src/lib/http/schema.d.ts generated vendored
View file

@ -25706,6 +25706,8 @@ export interface components {
input_cost_per_video_per_second_above_15s_interval?: number | null;
/** Input Cost Per Video Per Second Above 8S Interval */
input_cost_per_video_per_second_above_8s_interval?: number | null;
/** Input Cost Per Video Token */
input_cost_per_video_token?: number | null;
/** Itpm */
itpm?: number | null;
/** Litellm Credential Name */
@ -25787,6 +25789,8 @@ export interface components {
output_cost_per_token_priority?: number | null;
/** Output Cost Per Video Per Second */
output_cost_per_video_per_second?: number | null;
/** Output Cost Per Video Token */
output_cost_per_video_token?: number | null;
/** Output Vector Size */
output_vector_size?: number | null;
/** Quality Router Config */
@ -33536,6 +33540,8 @@ export interface components {
input_cost_per_video_per_second_above_15s_interval?: number | null;
/** Input Cost Per Video Per Second Above 8S Interval */
input_cost_per_video_per_second_above_8s_interval?: number | null;
/** Input Cost Per Video Token */
input_cost_per_video_token?: number | null;
/** Itpm */
itpm?: number | null;
/** Litellm Credential Name */
@ -33617,6 +33623,8 @@ export interface components {
output_cost_per_token_priority?: number | null;
/** Output Cost Per Video Per Second */
output_cost_per_video_per_second?: number | null;
/** Output Cost Per Video Token */
output_cost_per_video_token?: number | null;
/** Output Vector Size */
output_vector_size?: number | null;
/** Quality Router Config */