You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
# Reasoning family configurations (how to express reasoning for a family)
@@ -54,10 +54,10 @@ default_reasoning_effort: medium # low | medium | high
54
54
55
55
# Map concrete model names to a reasoning family
56
56
model_config:
57
-
"deepseek-v3":
57
+
"deepseek-v31":
58
58
reasoning_family: "deepseek"
59
59
preferred_endpoints: ["endpoint1"]
60
-
"qwen3-7b":
60
+
"qwen3-30b":
61
61
reasoning_family: "qwen3"
62
62
preferred_endpoints: ["endpoint1"]
63
63
"openai/gpt-oss-20b":
@@ -73,19 +73,11 @@ categories:
73
73
model_scores:
74
74
- model: openai/gpt-oss-20b
75
75
score: 1.0
76
-
- model: deepseek-v3
76
+
- model: deepseek-v31
77
77
score: 0.8
78
-
- model: qwen3-7b
78
+
- model: qwen3-30b
79
79
score: 0.8
80
80
81
-
- name: general
82
-
use_reasoning: false
83
-
reasoning_description: "General chit-chat doesn’t need reasoning"
84
-
model_scores:
85
-
- model: qwen3-7b
86
-
score: 1.0
87
-
- model: deepseek-v3
88
-
score: 0.8
89
81
90
82
# A safe default when no category is confidently selected
91
83
default_model: qwen3-7b
@@ -96,7 +88,8 @@ Notes
96
88
- A model only gets reasoning fields if it has a model_config.<MODEL>.reasoning_family that maps to a reasoning_families entry.
97
89
- DeepSeek/Qwen3 (chat_template_kwargs): the router injects chat_template_kwargs only when reasoning is enabled. When disabled, no chat_template_kwargs are added.
98
90
- GPT/GPT-OSS (reasoning_effort): when reasoning is enabled, the router sets reasoning_effort based on the category (fallback to default_reasoning_effort). When reasoning is disabled, if the request already contains reasoning_effort and the model’s family type is reasoning_effort, the router preserves the original value; otherwise it is absent.
99
-
- For more stable classification, you can add category descriptions in config and keep them semantically distinctive.
91
+
- Category descriptions (for example, description and reasoning_description) are informational only today; they do not affect routing or classification.
92
+
- Categories must be from MMLU-Pro at the moment; avoid free-form categories like "general". If you want generic categories, consider opening an issue to map them to MMLU-Pro.
0 commit comments