scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-08-28 05:25:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 36

fix(fireworks_ai): correct Kimi K2.5/K2.6/K2.7 max output token limits

Browse source 
Fireworks publishes a 262144-token context window for the Kimi K2.5, K2.6
and K2.7 models but caps generation well below that. Every fireworks_ai
Kimi K2.5/K2.6/K2.7 alias had max_output_tokens/max_tokens flattened to
262144 (equal to the context window), so the pre-call context-window check
admitted requests asking for a full 262144-token completion that Fireworks
rejects. Correct max_output_tokens/max_tokens to 32768 while keeping
max_input_tokens at 262144, and add a regression test pinning the limits
for all ten aliases.

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
mateo 2026-07-30 01:43:59 +00:00
parent 581f5c319e
commit f9c5be8ebf
3 changed files with 116 additions and 40 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

40
litellm/model_prices_and_context_window_backup.json
View file

@ -16703,8 +16703,8 @@
"input_cost_per_token": 6e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 3e-06,
"source": "https://fireworks.ai/pricing",
@ -16717,8 +16717,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -16733,8 +16733,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17077,8 +17077,8 @@
"input_cost_per_token": 6e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 3e-06,
"source": "https://fireworks.ai/pricing",
@ -17091,8 +17091,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17107,8 +17107,8 @@
"input_cost_per_token": 2e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17123,8 +17123,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17139,8 +17139,8 @@
"input_cost_per_token": 1.9e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -42501,8 +42501,8 @@
"input_cost_per_token": 2e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -42517,8 +42517,8 @@
"input_cost_per_token": 1.9e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",

40
model_prices_and_context_window.json
View file

@ -16703,8 +16703,8 @@
"input_cost_per_token": 6e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 3e-06,
"source": "https://fireworks.ai/pricing",
@ -16717,8 +16717,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -16733,8 +16733,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17077,8 +17077,8 @@
"input_cost_per_token": 6e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 3e-06,
"source": "https://fireworks.ai/pricing",
@ -17091,8 +17091,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17107,8 +17107,8 @@
"input_cost_per_token": 2e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17123,8 +17123,8 @@
"input_cost_per_token": 9.5e-07,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 4e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -17139,8 +17139,8 @@
"input_cost_per_token": 1.9e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -42622,8 +42622,8 @@
"input_cost_per_token": 2e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",
@ -42638,8 +42638,8 @@
"input_cost_per_token": 1.9e-06,
"litellm_provider": "fireworks_ai",
"max_input_tokens": 262144,
"max_output_tokens": 262144,
"max_tokens": 262144,
"max_output_tokens": 32768,
"max_tokens": 32768,
"mode": "chat",
"output_cost_per_token": 8e-06,
"source": "https://docs.fireworks.ai/serverless/pricing",

76
tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_kimi_model_metadata.py Normal file
View file

@ -0,0 +1,76 @@
"""
Regression test for Fireworks Kimi K2.5 / K2.6 / K2.7 context and output limits.
Fireworks publishes a 262144-token context window for every Kimi K2.5, K2.6 and
K2.7 model, but caps generation well below that. A previous bulk edit had flattened
max_output_tokens/max_tokens to 262144 (equal to the context window), which let the
pre-call context-window check admit requests asking for a full 262144-token
completion that Fireworks then rejects. These assertions pin the corrected per-alias
limits so a future bulk edit can't silently flatten them again.
"""
import json
from importlib.resources import files
import pytest
CONTEXT_WINDOW = 262144
OUTPUT_LIMIT = 32768
KIMI_ALIASES = (
"fireworks_ai/kimi-k2p5",
"fireworks_ai/kimi-k2p6",
"fireworks_ai/kimi-k2p6-fast",
"fireworks_ai/kimi-k2p7-code",
"fireworks_ai/kimi-k2p7-code-fast",
"fireworks_ai/accounts/fireworks/models/kimi-k2p5",
"fireworks_ai/accounts/fireworks/models/kimi-k2p6",
"fireworks_ai/accounts/fireworks/models/kimi-k2p7-code",
"fireworks_ai/accounts/fireworks/routers/kimi-k2p6-fast",
"fireworks_ai/accounts/fireworks/routers/kimi-k2p7-code-fast",
)
@pytest.fixture(scope="module")
def use_local_model_cost_map():
monkeypatch = pytest.MonkeyPatch()
monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True")
import litellm
from litellm.utils import _invalidate_model_cost_lowercase_map
original_model_cost = litellm.model_cost
litellm.model_cost = json.loads(
files("litellm")
.joinpath("model_prices_and_context_window_backup.json")
.read_text(encoding="utf-8")
)
litellm.get_model_info.cache_clear()
_invalidate_model_cost_lowercase_map()
try:
yield litellm
finally:
litellm.model_cost = original_model_cost
litellm.get_model_info.cache_clear()
_invalidate_model_cost_lowercase_map()
monkeypatch.undo()
@pytest.mark.parametrize("alias", KIMI_ALIASES)
def test_fireworks_kimi_raw_cost_entry_limits(use_local_model_cost_map, alias):
entry = use_local_model_cost_map.model_cost[alias]
assert entry["litellm_provider"] == "fireworks_ai"
assert entry["max_input_tokens"] == CONTEXT_WINDOW
assert entry["max_output_tokens"] == OUTPUT_LIMIT
assert entry["max_tokens"] == OUTPUT_LIMIT
assert entry["max_output_tokens"] < entry["max_input_tokens"]
@pytest.mark.parametrize("alias", KIMI_ALIASES)
def test_fireworks_kimi_get_model_info_limits(use_local_model_cost_map, alias):
model_info = use_local_model_cost_map.get_model_info(model=alias)
assert model_info["max_input_tokens"] == CONTEXT_WINDOW
assert model_info["max_output_tokens"] == OUTPUT_LIMIT
assert model_info["max_tokens"] == OUTPUT_LIMIT