What are the hardware requirements for CT LLM SFT DPO?

CT LLM SFT DPO requires a context window of 4K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed CT LLM SFT DPO and how large is it?

CT LLM SFT DPO is developed by m-a-p, a model with 2.1b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate CT LLM SFT DPO?

CT LLM SFT DPO is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

CT LLM SFT DPO by m-a-p — 4K context

Name: CT LLM SFT DPO
Author: m-a-p

CT LLM SFT DPO is an open-source language model by m-a-p. Features: 2.1b LLM, Context: 4K, License: apache-2.0, LLM Explorer Score: 0.12.

Arxiv:2404.04167 Endpoints compatible Llama Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/m-a-p/CT-LLM-SFT-DPO

CT LLM SFT DPO Parameters and Internals

Model Type

text-generation

Use Cases

Areas:

Academic research

Additional Notes

Developed for academic use, rigorous data compliance maintained. Disclaimer warns of potential liability for misuse.

Supported Languages

Chinese (High), English (Moderate)

Training Details

Data Sources:

cvalues_rlhf, comparison_gpt4_data_zh, oaast_rm_zh, huozi, zhihu, comparison_gpt4_data_en, beavertails, alpaca-gpt4, baichuan-6B

Data Volume:

183k Chinese pairs and 46k English pairs

Methodology:

Direct preference-based learning using DPO, aligned and enhanced over CT-LLM-SFT

Hardware Used:

8 H800 GPUs

Input Output

Input Format:

Chat conversation with user and system roles

Accepted Modalities:

text

Output Format:

Generated text responses

LLM Name	CT LLM SFT DPO
Repository 🤗	https://huggingface.co/m-a-p/CT-LLM-SFT-DPO
Model Size	2.1b
Required VRAM	0 GB
Updated	2026-07-17
Maintainer	m-a-p
Model Type	llama
Model Files	4.3 GB 0.0 GB 0.0 GB
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	4096
Model Max Length	4096
Transformers Version	4.37.2
Tokenizer Class	BaichuanTokenizer
Padding Token	<unk>
Vocabulary Size	125824
Torch Data Type	bfloat16

Rank the CT LLM SFT DPO capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to CT LLM SFT DPO

Best Alternatives	Context / RAM	Downloads	Likes
Sft	32K / 4.2 GB	5	0
Kanana 1.5 2.1B Instruct 2505	32K / 4.6 GB	5371	37
Fc Reasoning 2.1B	32K / 4.2 GB	8	0
Kanana 1.5 2.1B Base	32K / 4.2 GB	1534	13
Kanana Safeguard Prompt 2.1B	8K / 4.2 GB	1895	18
Kanana Nano 2.1B Instruct	8K / 4.2 GB	1025	75
Kanana Nano 2.1B Base	8K / 4.2 GB	391	40
DAMI Base Merge 0408	8K / 4.6 GB	7	0
EcomGen 0.0.2v	8K / 8.4 GB	8	0
CT LLM Base	4K / 4.3 GB	56	11

Note: green Score (e.g. "73.2") means that the model is better than m-a-p/CT-LLM-SFT-DPO.