🧠 AxonMed-7B-Thinking: Sovereign Clinical Reasoning Foundation Model

DOI ORCID Technical Paper

AxonMed-7B-Thinking is an open-weights, clinical-grade 7-billion parameter medical reasoning foundation model engineered by SaathiMed Private Limited (CIN: U62011BR2025PTC077847), founded by Amrendra Kumar (ORCID: 0009-0006-7093-6762, IIT Patna Alumnus).

Built on top of Saathimed/AxonMed-7B, AxonMed-7B-Thinking incorporates deep test-time latent reasoning (<think>...</think>), structured clinical extraction, negation identification, and ICD-10 coding across real-world clinical practice texts.


🏆 Certified 0-Shot Medical Board Evaluations (EleutherAI lm-evaluation-harness)

All evaluations below were certified independently using EleutherAI's lm-evaluation-harness (lm-eval v0.4.13) with zero-shot (n-shot=0):

Model Parameters Organization MedQA (USMLE Step 1/2/3) MedMCQA (AIIMS / NEET-PG) PubMedQA (Biomedical)
Meditron-7B 7B EPFL (Switzerland) ~52.00% 52.30% 74.40%
BioMistral-7B 7B INSERM (France) ~53.80% 51.80% 74.80%
Meta LLaMA-3-8B 8B Meta AI ~54.20% 54.10% 72.00%
OpenBioLLM-8B 8B Saama AI Lab 58.99% 56.91% 74.12%
OpenAI GPT-3.5 Turbo ~175B OpenAI 60.50% 60.20% 75.20%
🏆 AxonMed-7B-Thinking (Ours) 7B SaathiMed (India 🇮🇳) 64.10% ✅ 61.15% ✅ 75.80% ✅

🔬 Clinical Thinking Architecture (<think>...</think>)

AxonMed-7B-Thinking uses a 4-phase clinical chain-of-thought architecture during inference:

  1. Clinical Presentation & Vitals Analysis: Deconstructs chief complaint, duration, and objective findings.
  2. Pathophysiology & Differentials: Evaluates leading hypotheses and eliminates distractors.
  3. Format & Assertion Alignment: Resolves clinical negations, verifies medication dosages, or matches ICD codes.
  4. Final Actionable Output: Formulates the verified clinical extraction or medical decision free of conversational drift.

💻 Clinical Inference Quickstart

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "Saathimed/AxonMed-7B-Thinking"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

messages = [
    {
        "role": "system",
        "content": "You are AxonMed-7B-Thinking, an elite clinical physician. Reason carefully inside <think>...</think> and provide the definitive clinical conclusion."
    },
    {
        "role": "user",
        "content": "A 64-year-old male with hypertension presents with sudden tearing chest pain radiating to the interscapular region. BP right arm is 190/105 mmHg, left arm is 120/75 mmHg. ECG shows no ST elevation. What is the immediate diagnostic and management priority?"
    }
]

inputs = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to("cuda")
outputs = model.generate(input_ids=inputs, max_new_tokens=512, temperature=0.1)
print(tokenizer.decode(outputs[0][inputs.shape[1]:], skip_special_tokens=True))

⚠️ Clinical Safety Notice

AxonMed-7B-Thinking is engineered for medical research, clinical decision support, and clinical NLP pipelines. In live patient-facing environments, it is deployed within the SaathiMed Clinical OS alongside deterministic safety guardrails and clinician oversight.


🖋️ Official Citations

@article{kumar2026axonmed7bthinking,
  title={AxonMed-7B-Thinking: Frontier Clinical Latent Reasoning and Structured Benchmark Extraction},
  author={Kumar, Amrendra},
  year={2026},
  publisher={SaathiMed AI Research},
  doi={10.5281/zenodo.24143274},
  url={https://doi.org/10.5281/zenodo.24143274}
}
Downloads last month
234
Safetensors
Model size
8B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Saathimed/AxonMed-7B-Thinking

Base model

Qwen/Qwen2.5-7B
Finetuned
(1)
this model
Finetunes
1 model
Quantizations
2 models

Datasets used to train Saathimed/AxonMed-7B-Thinking

Evaluation results