GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
jithinodattu
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
jithinodattu
Model Tree
unsloth/Qwen3.5-2B
Input Modalities
Output Modalities
Supported Functionality
The model is intended for research and prototyping tasks such as:
Treat its output as analyst assistance. Verify technique IDs, claims, and recommended actions against the current MITRE ATT&CK knowledge base and the underlying evidence before operational use.
| Field | Value |
|---|---|
| Base model | unsloth/Qwen3.5-2B |
| Architecture | Qwen3_5ForConditionalGeneration |
| Parameters | approximately 2.27B |
| Modalities | text, image, and video inputs; text output |
| Native context length | 262,144 tokens |
| Weight format | Safetensors, full 16-bit checkpoint |
| Library | Transformers |
| License | Apache 2.0 |
The repository configuration was generated with Transformers 5.3.0. Use a recent Transformers release with Qwen3.5 support.
bash
pip install -U transformers accelerate torch pillowpython
from transformers import AutoModelForImageTextToText, AutoProcessor model_id = "jithinodattu/Qwen3.5-2B__mitreaware" processor = AutoProcessor.from_pretrained(model_id)model = AutoModelForImageTextToText.from_pretrained( model_id, torch_dtype="auto", device_map="auto",) messages = [ { "role": "user", "content": [ { "type": "text", "text": ( "Analyze the following observation and return the most likely " "MITRE ATT&CK technique IDs with brief evidence: " "PowerShell downloaded and executed an encoded payload." ), } ], }] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt",).to(model.device) generated = model.generate(**inputs, max_new_tokens=512)new_tokens = generated[:, inputs["input_ids"].shape[1]:]print(processor.batch_decode(new_tokens, skip_special_tokens=True)[0])For repeatable evaluation, prompt the model to return a fixed schema, for example:
json
{ "techniques": [ { "id": "TXXXX.XXX", "name": "Technique name", "evidence": "Text span or observation supporting the mapping", "confidence": 0.0 } ]}This checkpoint was fine-tuned with Unsloth and Hugging Face TRL. The training dataset, preprocessing procedure, hyperparameters, compute configuration, and task-specific evaluation results have not yet been documented in this model card. Until those details and benchmark results are published, users should not assume a measured improvement over the base model.
Recommended evaluation should include technique-level precision, recall, and F1; exact and hierarchical ATT&CK matching; hallucinated technique rate; and performance on reports from sources not represented in the training data.
This checkpoint is released under the Apache 2.0 license. It is derived from
unsloth/Qwen3.5-2B, which in turn
derives from Qwen3.5-2B. Review the upstream model card for additional base-model
details, usage guidance, and citations.
Fine-tuning and checkpoint publication: jithinodattu.
pip install -U transformers accelerate torch pillowfrom transformers import AutoModelForImageTextToText, AutoProcessor model_id = "jithinodattu/Qwen3.5-2B__mitreaware" processor = AutoProcessor.from_pretrained(model_id)model = AutoModelForImageTextToText.from_pretrained( model_id, torch_dtype="auto", device_map="auto",) messages = [ { "role": "user", "content": [ { "type": "text", "text": ( "Analyze the following observation and return the most likely " "MITRE ATT&CK technique IDs with brief evidence: " "PowerShell downloaded and executed an encoded payload." ), } ], }] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt",).to(model.device) generated = model.generate(**inputs, max_new_tokens=512)new_tokens = generated[:, inputs["input_ids"].shape[1]:]print(processor.batch_decode(new_tokens, skip_special_tokens=True)[0]){ "techniques": [ { "id": "TXXXX.XXX", "name": "Technique name", "evidence": "Text span or observation supporting the mapping", "confidence": 0.0 } ]}