NTele-Hybrid-27B-v1

One model for generation and decision tasks.

NTele-Hybrid-27B-v1, fine-tuned from Qwen3.8-27B, is a 27B hybrid model combining text generation with structured decision-making. Deploy a single model instance and access both capabilities through separate API endpoints.

Performance

Decision Index

Evaluated on the Decision Index 0.3 Public suite.

Model	Public Index	Knowledge & Reasoning	Language Understanding	Retrieval & Classification	Tools & Automation	Arts & Human Taste
Qwen3.8-27B	50.26	32.31	53.25	55.22	73.24	36.95
Jev API 1.14 (official reference)	60.70	57.12	61.18	63.53	73.78	39.16
NTele-Hybrid-27B-v1	63.25	45.62	70.49	64.34	79.69	57.89

Jev API 1.14 results come from the official public snapshot.

Generation tasks

NTele-Hybrid-27B-v1 matches or exceeds the base model across all five evaluations below.

Model	LiveCodeBench-V6	GPQA Diamond	MATH500	IFEval	AIME2025
Qwen3.5-27B	59.14	65.15	92.40	85.71	36.67
NTele-Hybrid-27B-v1	59.50	68.18	93.20	87.34	40.00

Quick start

We recommend using SGLang ≥ 0.5.20 to deploy the model.

python -m sglang.launch_server \
  --model-path /path/to/NTele-Hybrid-27B-v1 \
  --served-model-name NTele-Hybrid-27B-v1 \
  --host 0.0.0.0 \
  --port 7801

Generation

curl http://localhost:7801/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "NTele-Hybrid-27B-v1",
    "messages": [
      {"role": "user", "content": "Explain binary search."}
    ]
  }'

Decision

Use the calibrated temperature 1.26964 for decision tasks.

curl -X POST "http://localhost:7801/v1/decisions" \
  -H "Content-Type: application/json" \
  --max-time 60 \
  --data-raw '{
    "input": "I have been trying to connect my Stripe account for 3 days and the integration keeps failing. I am losing sales.",
    "temperature": 1.26964,
    "prompt_format_version": 1,
    "return_prompt_token_ids": true,
    "questions": [
      {
        "id": "team",
        "type": "choice",
        "question": "Which team should handle this ticket?",
        "options": [
          { "name": "billing", "description": "Payment or subscription issues" },
          { "name": "technical", "description": "Bugs or integration problems" },
          { "name": "sales", "description": "Pricing or account questions" }
        ]
      },
      {
        "id": "frustration",
        "type": "score",
        "question": "How frustrated is the customer?",
        "levels": ["Calm", "Frustrated but civil", "Very angry"]
      },
      {
        "id": "urgent",
        "type": "yes_no",
        "question": "The customer needs an answer today."
      }
    ]
  }'