Skip to content
Computer vision

TCM-Qwen3.5-9B-Instruct-CPT-1029

a model for medical text analysis

2 min readupdated

This is a specialized version of Qwen/Qwen3.5-9B, fine-tuned via continuous pre-training (CPT) on medical texts, textbooks, and patient records. It merges the base architecture weights with a language LoRA adapter while preserving chat capabilities. Built for research in medical text processing and assisting with specialized literature analysis, the model is not intended for autonomous diagnosis or prescribing treatment without medical professionals.

Key features

  • Medical domain adaptation: Trained on a dataset of medical books, clinical knowledge, and case histories.
  • Lower PPL: Perplexity (a measure of text uncertainty) on medical data dropped by over 33%.
  • Benchmark gains: Shows improved scores across several tasks, including the C-Eval exam benchmark.
  • Merged weights: The adapter is already merged into the base model, making it easy to run without extra libraries.

What you can do

  • Literature review: Summarize and extract information from specialized medical sources.
  • Chatbot integration: Set up a local assistant to answer questions on medical theory.
  • Output comparison: Compare how it generates medical terminology against the base Qwen model.
  • Text processing: Extract entities from medical reports and clinical notes.

Example

from transformers import AutoModelForImageTextToText, AutoProcessor

model_id = "Xxiao2023/TCM-Qwen3.5-9B-Instruct-CPT-1029"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForImageTextToText.from_pretrained(model_id, device_map="auto")
Read moreDetails
Based on Qwen/Qwen3.5-9B with 9.41B parameters.
Benchmark results: C-Eval rose from 61.44% to 65.53%, HellaSwag reached 70.86%.
Trained on a single NVIDIA RTX A6000 (48GB).
Requires a compatible environment with Hugging Face Transformers 5.17.0 and PyTorch.
Discuss Telegram 0 13

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 18 GB · ready files20 GB
  • FP8 / INT8recommendedweights 8.8 GB10 GB
  • 4-bitweights 4.9 GB5.9 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2080 Ti · BF16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Xxiao2023/TCM-Qwen3.5-9B-Instruct-CPT-1029
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Xxiao2023/TCM-Qwen3.5-9B-Instruct-CPT-1029 --local-dir ~/models/TCM-Qwen3.5-9B-Instruct-CPT-1029
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

  1. Qwen3.5-9B-Basefamily root
  2. Qwen3.5-9B
  3. TCM-Qwen3.5-9B-Instruct-CPT-1029

New posts about versions of Qwen3.5-9B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 6All on Hugging Face
Merges 1All on Hugging Face

About the model

Xxiao2023/TCM-Qwen3.5-9B-Instruct-CPT-1029
Author
Xxiao2023
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

130

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Audio

Qwen3-ASR

Qwen3-ASR
Robotics

Jevol-2B

Jevol-2B