# Open-Weight Models Under llama3.2: Open-Weight AI Source: https://savrn.com/models/licenses/llama3-2 Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- SAVRN Model Hub · Models by License # Open-Weight Models Under llama3.2 40 open-weight models released under llama3.2 in the SAVRN Model Hub, with Yanggangu, Varun Sampath Kumar and Meta Llama publishing the most. 40Models 19Publishers 750M to 3.9BParameter range 1Licenses ## SAVRN's Take We do not summarize what the llama3.2 license allows for commercial use. The license text governs, and it is what a buyer's counsel reads before anyone draws a deployment plan. Five models carry it in our catalog, all text generation. The three from Meta Llama are gated, so an organization clears the publisher's access step before any weights reach its facility. The two derivatives, a medical fine-tune quantized by TinyTech and a logistics checkpoint from Yuchi Wang, are open access yet carry the same license, so a downstream fine-tune does not change the terms you operate under. A 1.2B parameter model needs 3.0 GB at 16-bit, 1.5 GB at 8-bit and 0.7 GB at 4-bit; the 3.2B Instruct needs 7.7 GB, 3.9 GB and 1.9 GB. The cheapest Index host for every one of those figures is a single MI300X at $1.85 an hour, so quantizing does not lower the rate, it lowers the share of that card one instance occupies. The DelaySentinel checkpoint lists a context length of 131,072 tokens; the Meta Llama entries list none, so confirm that figure with the publisher before sizing long inputs. Demand concentrates at the small end. Llama-3.2-1B-Instruct drew 6,921,311 downloads in a month, Llama-3.2-3B-Instruct 1,859,982, and the base Llama-3.2-1B 905,658. The TinyTech medical GGUF sits at 1,568 and DelaySentinel at 42. Serve customers and the 1B Instruct is the one they have most likely already run. SAVRN Research, 2026-09-18 ## Most Downloaded | Model | Publisher | Parameters | License | Monthly downloads | Cheapest GPUs at 16-bit | | --- | --- | --- | --- | --- | --- | | [Llama-3.2-1B-Instruct](https://savrn.com/models/llama-3-2-1b-instruct) | Meta Llama | 1.2B | llama3.2 | 4M | 1x MI300X, $1.85/hr | | [Llama-3.2-3B-Instruct](https://savrn.com/models/llama-3-2-3b-instruct) | Meta Llama | 3.2B | llama3.2 | 1.1M | 1x MI300X, $1.85/hr | | [Llama-3.2-1B](https://savrn.com/models/llama-3-2-1b) | Meta Llama | 1.2B | llama3.2 | 905.7k | 1x MI300X, $1.85/hr | | [Llama-3.2-3B-Instruct-Medical-Finetuned-merged-GGUF](https://savrn.com/models/llama-3-2-3b-instruct-medical-finetuned-merged-gguf) | TinyTech | — | llama3.2 | 1.6k | — | | [Llama-3.2-1B-dolma3_mix](https://savrn.com/models/llama-3-2-1b-dolma3-mix) | CHEN EN LO | 1.2B | llama3.2 | 523 | 1x MI300X, $1.85/hr | | [Llama-3.2-1B-Instruct-ExecuTorch](https://savrn.com/models/llama-3-2-1b-instruct-executorch) | Experimental Machines | — | llama3.2 | 264 | — | | [Mina-MoE-4x1B-RL-Zero](https://savrn.com/models/mina-moe-4x1b-rl-zero) | Bùi Minh Trí | 3.9B | llama3.2 | 201 | 1x MI300X, $1.85/hr | | [StockLLM](https://savrn.com/models/stockllm) | The Fin AI | 1.2B | llama3.2 | 178 | 1x MI300X, $1.85/hr | | [Llama-3.2-3B-Instruct-RKLLM](https://savrn.com/models/llama-3-2-3b-instruct-rkllm) | Hanzo Huang | — | llama3.2 | 44 | — | | [Llama-3.2-1B-DelaySentinel](https://savrn.com/models/llama-3-2-1b-delaysentinel) | Yuchi Wang | 1.2B | llama3.2 | 42 | 1x MI300X, $1.85/hr | ## Who Publishes Them | Publisher | Models | | --- | --- | | [Yanggangu](https://savrn.com/model-publishers/yanggangu) | 14 | | [Varun Sampath Kumar](https://savrn.com/model-publishers/lordvarun23) | 4 | | [Meta Llama](https://savrn.com/model-publishers/meta-llama) | 3 | | [Yothin](https://savrn.com/model-publishers/yothins) | 2 | | [Hanzo Huang](https://savrn.com/model-publishers/hanzohuang) | 2 | | [Haoran Zheng](https://savrn.com/model-publishers/flyingodzilla) | 2 | ## All 40 Models Model · Text generation ### [Llama-3.2-1B-Instruct](https://savrn.com/models/llama-3-2-1b-instruct) [Meta Llama](https://savrn.com/model-publishers/meta-llama) The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out). The Llama 3.2 instruction-tuned text only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarization tasks. They outperform many of the available open source and closed chat models on common industry benchmarks. Model Architecture: Llama 3.2 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for… Access requested at publisher llama3.2 1.2B parameters transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct) Model · Text generation ### [Llama-3.2-3B-Instruct](https://savrn.com/models/llama-3-2-3b-instruct) [Meta Llama](https://savrn.com/model-publishers/meta-llama) The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out). The Llama 3.2 instruction-tuned text only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarization tasks. They outperform many of the available open source and closed chat models on common industry benchmarks. Model Architecture: Llama 3.2 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for… Access requested at publisher llama3.2 3.2B parameters transformers [View model](https://savrn.com/models/llama-3-2-3b-instruct) Model · Text generation ### [Llama-3.2-1B](https://savrn.com/models/llama-3-2-1b) [Meta Llama](https://savrn.com/model-publishers/meta-llama) The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out). The Llama 3.2 instruction-tuned text only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarization tasks. They outperform many of the available open source and closed chat models on common industry benchmarks. Model Architecture: Llama 3.2 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for… Access requested at publisher llama3.2 1.2B parameters transformers [View model](https://savrn.com/models/llama-3-2-1b) Model · Image and text to text ### [StockLLM](https://savrn.com/models/stockllm) [The Fin AI](https://savrn.com/model-publishers/thefinai) StockLLM is an open-source fine-tuned 1B large language model as the backbone of our first retrieval-augmented generation (RAG) framework specifically designed for financial time-series forecasting. Paper or resources for more information: https://arxiv.org/pdf/2502.05878 This repository and its contents are provided for academic and educational purposes only. None of the material constitutes financial, legal, or investment advice. No warranties, express or implied, are offered regarding the accuracy, completeness, or utility of the content. The authors and contributors are not responsible for any errors, omissions, or any consequences arising from the use of the information herein. Users… Access requested at publisher llama3.2 1.2B parameters transformers [View model](https://savrn.com/models/stockllm) Model · Text generation ### [Llama-3.2-1B-DelaySentinel](https://savrn.com/models/llama-3-2-1b-delaysentinel) [Yuchi Wang](https://savrn.com/model-publishers/yuchiwang02) Building and evaluating AI on logistics data. Built with Llama. This checkpoint is a full-parameter fine-tune of meta-llama/Llama-3.2-1B-Instruct, published with the data split, evaluation results, and behavioral tests used to inspect it. The project connects a working training-and-inference pipeline with a retrospective audit of what its score demonstrates. The checkpoint and a depth-2 decision tree both score 100% on the same historical 200-row split. A rule using two supplied fields reproduces every label in the 1,000-row source table. Prompt rewrites reveal additional response failures, including sensitivity to irrelevant text. These findings make the checkpoint useful for studying… Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-delaysentinel) Model · Text generation ### [Llama-3.2-3B-Instruct-Medical-Finetuned-merged-GGUF](https://savrn.com/models/llama-3-2-3b-instruct-medical-finetuned-merged-gguf) [TinyTech](https://savrn.com/model-publishers/tinyopsec) GGUF quantizations of MohamedAhmedAE/Llama-3.2-3B-Instruct-Medical-Finetuned-merged — a 3.21B-parameter Llama-3.2 model adapted to the medical domain using QLoRA, with the LoRA weights merged back into the base. Quantized by tinyopsec. Med-LLaMA3.2-3B is the 3B (balanced / mid-tier) member of the Med-LLaMA3 family introduced in the paper "Med-LLaMA3: Advancing Medical Question-Answering Through Parameter-Efficient Fine-Tuning of Large Language Models" (Applied Sciences, 2026). The family adapts the LLaMA-3 architecture to the medical domain by training only a small fraction of the base model's parameters (5.70% for this 3B variant), achieving strong medical question-answering performance… Open weights llama3.2 gguf [View model](https://savrn.com/models/llama-3-2-3b-instruct-medical-finetuned-merged-gguf) Model · Text generation ### [Llama-3.2-1B-dolma3_mix](https://savrn.com/models/llama-3-2-1b-dolma3-mix) [CHEN EN LO](https://savrn.com/model-publishers/asciibase64) This revision publishes the final raw checkpoint of raw6b-normuon-adam2x-20261018: 6,000,082,944 non-padding input tokens, 28,619 optimizer updates. The model was randomly initialized (residual-half, seed42); no pretrained weights or previous training checkpoint initialized this run. Model architecture and The full reserved 50,000-document set has not been evaluated for this checkpoint. PPL is exp(sum held-out NLL / sum valid prediction tokens), not a mean of subset PPLs. Documents are evaluated independently, with stride8191 for long documents. The entire original document split was shuffled with seed42 before tokenization; the last50,000 documents were held out and excluded from all… Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-dolma3-mix) Model · Text generation ### [Llama-3.2-1B-Instruct-ExecuTorch](https://savrn.com/models/llama-3-2-1b-instruct-executorch) [Experimental Machines](https://savrn.com/model-publishers/experimentalmachines) Built with Llama ExecuTorch exports of meta-llama/Llama-3.2-1B-Instruct (revision 9213176726f5) for on-device inference with the openweights Android app or any ExecuTorch 1.4.0 runtime. Windows in this repo: XNNPACK (CPU) at 2k to 32k; Vulkan (GPU) at 2k to 32k; Qualcomm QNN (HTP) SM8750 (Snapdragon 8 Elite) at 2k to 8k; MediaTek NeuroPilot MT6991 (Dimensity 9400) at 2k to 16k. The window is fixed inside the file: the runtime allocates the whole KV cache at load, so pick the largest window the device can hold (fitsphonebudget in each folder's config.json is the estimate against a 5 GB budget). MediaTek folders also hold the token embedding table the NeuroPilot runner reads from disk, shared… Open weights llama3.2 executorch [View model](https://savrn.com/models/llama-3-2-1b-instruct-executorch) Model ### [Mina-MoE-4x1B-RL-Zero](https://savrn.com/models/mina-moe-4x1b-rl-zero) [Bùi Minh Trí](https://savrn.com/model-publishers/trivnpk) Mô hình ngôn ngữ nhỏ phân hóa 4 Chuyên gia (Sparse Mixture of Experts) theo kiến trúc Mixtral 4x1B. - Tổng tham số lưu trữ: 3,651,864,576 (~3.65B) - Tham số kích hoạt mỗi token: 1,235,879,936 (1.236B Active FLOPs) - Phiên bản GGUF: Q6K (~2.4 GB) sweet spot, Q80 (~3.8 GB) benchmark Open weights llama3.2 3.9B parameters 131,072 tokens [View model](https://savrn.com/models/mina-moe-4x1b-rl-zero) Model · Text generation ### [Llama-3.2-3B-Instruct-RKLLM](https://savrn.com/models/llama-3-2-3b-instruct-rkllm) [Hanzo Huang](https://savrn.com/model-publishers/hanzohuang) Text generation on Rockchip RK3576 and RK3588, converted from meta-llama/Llama-3.2-3B-Instruct. Use the files for your board’s exact SoC. w4a16 uses 4-bit weights and 16-bit activations; w4a16g128 adds a weight group size of 128. w8a8 uses 8-bit weights and activations. File size below is download size, not total runtime RAM. With Docker installed on a matching Rockchip board, this example starts the RK3576 w4a16g128 model and its API on port 8001: The image includes its model file. See rkllm-docker for API requests, other board variants, and configuration. To download the files separately instead: These are native RKLLM/RKNN artifacts. Use a compatible RKLLM runtime and the model’s correct… Open weights llama3.2 rkllm [View model](https://savrn.com/models/llama-3-2-3b-instruct-rkllm) Model · Text generation ### [Llama-3.2-1B-Instruct-RKLLM](https://savrn.com/models/llama-3-2-1b-instruct-rkllm) [Hanzo Huang](https://savrn.com/model-publishers/hanzohuang) Text generation on Rockchip RK3576 and RK3588, converted from meta-llama/Llama-3.2-1B-Instruct. Use the files for your board’s exact SoC. w4a16 uses 4-bit weights and 16-bit activations; w4a16g128 adds a weight group size of 128. w8a8 uses 8-bit weights and activations. File size below is download size, not total runtime RAM. With Docker installed on a matching Rockchip board, this example starts the RK3576 w4a16g128 model and its API on port 8001: The image includes its model file. See rkllm-docker for API requests, other board variants, and configuration. To download the files separately instead: These are native RKLLM/RKNN artifacts. Use a compatible RKLLM runtime and the model’s correct… Open weights llama3.2 rkllm [View model](https://savrn.com/models/llama-3-2-1b-instruct-rkllm) Model · Text generation ### [iris-llama-3.2-3b-combined-multihop-k10-ep4](https://savrn.com/models/iris-llama-3-2-3b-combined-multihop-k10-ep4) [Darya Zare](https://savrn.com/model-publishers/daryazare) This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set: - Info Ndcg@p8: 0.9770 The following hyperparameters were used during training: - learningrate: 5e-05 - trainbatchsize: 1 - evalbatchsize: 1 - gradientaccumulationsteps: 16 - totaltrainbatchsize: 16 - lrschedulertype: constant - lrschedulerwarmupsteps: 0.03 - numepochs: 4.0 - PEFT 0.19.1 - Transformers 5.7.0 - Pytorch 2.11.0+cu130 - Datasets 4.8.5 - Tokenizers 0.22.2 Open weights llama3.2 peft [View model](https://savrn.com/models/iris-llama-3-2-3b-combined-multihop-k10-ep4) Model · Text generation ### [Llama-3.2-1B-Instruct-Conceptizer](https://savrn.com/models/llama-3-2-1b-instruct-conceptizer) [Haoran Zheng](https://savrn.com/model-publishers/flyingodzilla) Open weights llama3.2 peft [View model](https://savrn.com/models/llama-3-2-1b-instruct-conceptizer) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-20Minuten](https://savrn.com/models/llama-3-2-1b-instruct-ft-20minuten) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-20minuten) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-C-STANCE](https://savrn.com/models/llama-3-2-1b-instruct-ft-c-stance) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-c-stance) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-FOMC](https://savrn.com/models/llama-3-2-1b-instruct-ft-fomc) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-fomc) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-MeetingBank](https://savrn.com/models/llama-3-2-1b-instruct-ft-meetingbank) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-meetingbank) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-NumGLUE-cm](https://savrn.com/models/llama-3-2-1b-instruct-ft-numglue-cm) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-numglue-cm) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-NumGLUE-ds](https://savrn.com/models/llama-3-2-1b-instruct-ft-numglue-ds) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-numglue-ds) Model · Text generation ### [Llama-3.2-1B-Instruct-FT-ScienceQA](https://savrn.com/models/llama-3-2-1b-instruct-ft-scienceqa) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A FT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-ft-scienceqa) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-20Minuten](https://savrn.com/models/llama-3-2-1b-instruct-smat-20minuten) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-20minuten) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-C-STANCE](https://savrn.com/models/llama-3-2-1b-instruct-smat-c-stance) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-c-stance) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-FOMC](https://savrn.com/models/llama-3-2-1b-instruct-smat-fomc) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-fomc) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-MeetingBank](https://savrn.com/models/llama-3-2-1b-instruct-smat-meetingbank) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-meetingbank) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-NumGLUE-cm](https://savrn.com/models/llama-3-2-1b-instruct-smat-numglue-cm) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-numglue-cm) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-NumGLUE-ds](https://savrn.com/models/llama-3-2-1b-instruct-smat-numglue-ds) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-numglue-ds) Model · Text generation ### [Llama-3.2-1B-Instruct-SMAT-ScienceQA](https://savrn.com/models/llama-3-2-1b-instruct-smat-scienceqa) [Yanggangu](https://savrn.com/model-publishers/yanggangu) A SMAT expert from Table 1 of SMAT: Simple and Efficient Merge-Aware Training (training seed 42). SMAT trains experts that retain task performance while improving model merging. Built with Llama. The included Llama 3.2 license and acceptable-use policy apply. Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-smat-scienceqa) Model · Text generation ### [Llama-3.2-1B-Instruct-TOFU-forget10-GradDiff](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-graddiff) [Varun Sampath Kumar](https://savrn.com/model-publishers/lordvarun23) open-unlearning/tofuLlama-3.2-1B-Instructfull (Llama-3.2-1B-Instruct fine-tuned on all 4,000 TOFU QA pairs) after unlearning the forget10 split (20 fictitious authors) with GradDiff. GradDiff (Gradient Difference; Liu et al., 2022; OpenUnlearning implementation): gradient ascent on the forget set + NLL on the retain set: $\mathcal{L} = -\gamma\,\mathrm{NLL}(\text{forget}) + \alpha\,\mathrm{NLL}(\text{retain})$. TOFU contains 200 fictitious authors with 20 QA pairs each. forget10 holds 10% of them (20 authors, 400 QA pairs); retain90 holds the other 180 authors. The oracle is the same base model fine-tuned on retain90 only, i.e. a model that never saw the forget set. Hyperparameter… Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-graddiff) Model · Text generation ### [Llama-3.2-1B-Instruct-TOFU-forget10-GradDiff-ILU](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-graddiff-ilu) [Varun Sampath Kumar](https://savrn.com/model-publishers/lordvarun23) open-unlearning/tofuLlama-3.2-1B-Instructfull (Llama-3.2-1B-Instruct fine-tuned on all 4,000 TOFU QA pairs) after unlearning the forget10 split (20 fictitious authors) with GradDiff + ILU. GradDiff (Gradient Difference; Liu et al., 2022; OpenUnlearning implementation): gradient ascent on the forget set + NLL on the retain set: $\mathcal{L} = -\gamma\,\mathrm{NLL}(\text{forget}) + \alpha\,\mathrm{NLL}(\text{retain})$. ILU (Invariant LLM Unlearning; Wang et al., ICML 2025) is applied on top of GradDiff as a relearning defense: adds an IRMv1 invariance penalty computed on an unrelated task (GSM8K) so that the unlearning solution stays stationary under fine-tuning on other data (reimplemented… Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-graddiff-ilu) Model · Text generation ### [Llama-3.2-1B-Instruct-TOFU-forget10-GradDiff-SAM](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-graddiff-sam) [Varun Sampath Kumar](https://savrn.com/model-publishers/lordvarun23) open-unlearning/tofuLlama-3.2-1B-Instructfull (Llama-3.2-1B-Instruct fine-tuned on all 4,000 TOFU QA pairs) after unlearning the forget10 split (20 fictitious authors) with GradDiff + SAM. GradDiff (Gradient Difference; Liu et al., 2022; OpenUnlearning implementation): gradient ascent on the forget set + NLL on the retain set: $\mathcal{L} = -\gamma\,\mathrm{NLL}(\text{forget}) + \alpha\,\mathrm{NLL}(\text{retain})$. SAM (Sharpness-Aware Minimization for robust unlearning; Fan et al., ICML 2025) is applied on top of GradDiff as a relearning defense: the forget loss is evaluated at the sharpness-aware perturbed weights $\theta + \rho\,g/\lVert g\rVert$ (retain gradient at $\theta$), so the… Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-graddiff-sam) Model · Text generation ### [Llama-3.2-1B-Instruct-TOFU-forget10-NPO](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-npo) [Varun Sampath Kumar](https://savrn.com/model-publishers/lordvarun23) open-unlearning/tofuLlama-3.2-1B-Instructfull (Llama-3.2-1B-Instruct fine-tuned on all 4,000 TOFU QA pairs) after unlearning the forget10 split (20 fictitious authors) with NPO. NPO (Negative Preference Optimization; Zhang et al., 2024): $\mathcal{L} = \gamma\,\tfrac{2}{\beta}\,\mathbb{E}{\text{forget}}\big[\log\big(1+(\pi\theta/\pi{\text{ref}})^{\beta}\big)\big] + \alpha\,\mathrm{NLL}(\text{retain})$, with $\pi{\text{ref}}$ the frozen starting model. TOFU contains 200 fictitious authors with 20 QA pairs each. forget10 holds 10% of them (20 authors, 400 QA pairs); retain90 holds the other 180 authors. The oracle is the same base model fine-tuned on retain90 only, i.e. a model that never saw… Open weights llama3.2 1.2B parameters 131,072 tokens transformers [View model](https://savrn.com/models/llama-3-2-1b-instruct-tofu-forget10-npo) Model · Text generation ### [Llama-3.2-3B-Instruct-Conceptizer](https://savrn.com/models/llama-3-2-3b-instruct-conceptizer) [Haoran Zheng](https://savrn.com/model-publishers/flyingodzilla) Open weights llama3.2 peft [View model](https://savrn.com/models/llama-3-2-3b-instruct-conceptizer) Model ### [Llama-3.2-3B-Instruct-DSpark](https://savrn.com/models/llama-3-2-3b-instruct-dspark) [Yosef Worku Alemneh](https://savrn.com/model-publishers/rasyosef) A DSpark draft model for speculative decoding with unsloth/Llama-3.2-3B-Instruct as the verifier, trained with speculators. The drafter proposes 8 tokens at a time and the verifier checks them in one forward pass, so output is identical to running the verifier alone — a lossless speedup. Mean acceptance length is 3.27 tokens committed per verification step, up to 4.92 on mathreasoning. vLLM loads the verifier automatically from the config — don't pass it separately. Then query the OpenAI-compatible endpoint at http://localhost:8000/v1. 5 Qwen3 layers (hidden size 3072, intermediate size 8192, 24 attention heads over 8 KV heads, sliding-window attention with a 2048-token window), ~0.9B… Open weights llama3.2 750M parameters transformers [View model](https://savrn.com/models/llama-3-2-3b-instruct-dspark) Model · Text generation ### [Llama-3.2-3B-SPARQL-GGUF](https://savrn.com/models/llama-3-2-3b-sparql-gguf) [Duban Potosi](https://savrn.com/model-publishers/dubansmith) Built with Llama Modelo basado en meta-llama/Llama-3.2-3B-Instruct, ajustado con LoRA para generar consultas SPARQL. Modelo base con el adaptador LoRA: Describe aquí tu dataset, hiperparámetros y métricas. Distribuido bajo la Llama 3.2 Community License. Ver LICENSE y NOTICE. Uso sujeto a la Acceptable Use Policy de Meta. Open weights llama3.2 [View model](https://savrn.com/models/llama-3-2-3b-sparql-gguf) Model · Text generation ### [Llama-3.2-3B-Thai_CRM](https://savrn.com/models/llama-3-2-3b-thai-crm) [Yothin](https://savrn.com/model-publishers/yothins) Fine-tuned version of meta-llama/Llama-3.2-3B optimized for Customer Relationship Management (CRM), Customer Behavior Analysis, and Thai business workflows. Trained using Unsloth and Hugging Face's TRL library. from transformers import AutoModelForCausalLM, AutoTokenizer import torch modelid = "yothinS/Llama-3.2-3B-ThaiCRM" tokenizer = AutoTokenizer.frompretrained(modelid) model = AutoModelForCausalLM.frompretrained( modelid, torchdtype=torch.bfloat16, devicemap="auto" messages = [ {"role": "system", "content": "You are a professional CRM and customer behavior specialist."}, {"role": "user", "content": "วิเคราะห์พฤติกรรมลูกค้ารายนี้และแนะนำโปรโมชันรักษาฐานลูกค้า: ยอดซื้อเฉลี่ยลดลง 40%… Open weights llama3.2 3.2B parameters 131,072 tokens [View model](https://savrn.com/models/llama-3-2-3b-thai-crm) Model · Text generation ### [Llama-3.2-3B-Thai-CRM-GGUF](https://savrn.com/models/llama-3-2-3b-thai-crm-gguf) [Yothin](https://savrn.com/model-publishers/yothins) This repository contains GGUF quantized weights for yothinS/Llama-3.2-3B-ThaiCRM, a model fine-tuned for Customer Relationship Management (CRM) workflows, Thai business communications, and customer behavioral analysis. GGUF files can be run locally on CPU, Apple Silicon, or NVIDIA/AMD GPUs via llama.cpp, Ollama, LM Studio, or carrany/text-generation-webui. Download your chosen GGUF file using the Hugging Face CLI: huggingface-cli download yothinS/Llama-3.2-3B-ThaiCRM-GGUF Llama-3.2-3B-ThaiCRM.Q4KM.gguf --local-dir. Open weights llama3.2 [View model](https://savrn.com/models/llama-3-2-3b-thai-crm-gguf) Model · Reinforcement learning ### [Llama-Qwen-few-gt-transfer](https://savrn.com/models/llama-qwen-few-gt-transfer) [Jae Woo Shin](https://savrn.com/model-publishers/jaewooshin) 96 for Qwen and 128 for Llama, compared with 32 for both in v4. Full runs retain 64 prompts, 16 responses per prompt, and 300 total steps. The 24 GB offload profile retains concurrency 8. See the v5 release notes. Higher concurrency has CPU configuration validation only; GPU capacity and throughput have not been measured, and sampled training trajectories may differ. The guided launcher uses majority voting; wrong-label and inverted reward selection are not exposed by this launcher. V5 includes the v4 fix for caches creating a run directory before the launcher. For the standalone v3 cache fix, see the patch notes. Use a separate extraction folder when upgrading. Existing model, checkpoint… Open weights llama3.2 pytorch [View model](https://savrn.com/models/llama-qwen-few-gt-transfer) Model ### [manifold-1.0-1b](https://savrn.com/models/manifold-1-0-1b) [Naveen Mysore](https://savrn.com/model-publishers/nmysore) Attention manifold checkpoint for LLaMA 3.2 1B Instruct. This model extends the standard transformer attention mechanism with learned manifold structures on the value path. The checkpoint contains the manifold parameters only — the base model weights are loaded from meta-llama/Llama-3.2-1B-Instruct. See the LLM Inspector for the full visualization and editing toolkit. Paper forthcoming. Open weights llama3.2 [View model](https://savrn.com/models/manifold-1-0-1b) Model ### [MeLlamFox](https://savrn.com/models/mellamfox) [Micro Fox](https://savrn.com/model-publishers/mircofox-ai) basemodel: - meta-llama/Llama-3.2-1B-Instruct libraryname: transformers - llama3.2 Open weights llama3.2 transformers [View model](https://savrn.com/models/mellamfox) Model · Text to audio ### [MIDI-LLM_Llama-3.2-1B-GGUF](https://savrn.com/models/midi-llm-llama-3-2-1b-gguf) [Trinton Azaleth](https://savrn.com/model-publishers/tazaleth) Built with Llama. Quantized GGUF build of slseanwu/MIDI-LLMLlama-3.2-1B for local inference in MIDI Forge via llama.cpp. - Do not use truncated GGUFs that drop the MIDI embedding rows. This model is a derivative of Meta Llama 3.2 and is subject to the and Acceptable Use Policy. 1. Creative → Generate with MIDI-LLM… 2. Accept the Llama 3.2 license in the dialog 3. Download model (this file) or the app fetches it automatically from: https://huggingface.co/tazaleth/MIDI-LLMLlama-3.2-1B-GGUF/resolve/main/midi-llm-llama32-1b-q5km.gguf AMT event vocabulary: Thickstun et al., Anticipatory Music Transformer (TMLR 2024). Open weights llama3.2 gguf [View model](https://savrn.com/models/midi-llm-llama-3-2-1b-gguf) ## Questions ### Which llama3.2 models are most downloaded? By monthly downloads reported by the Hugging Face Hub: Llama-3.2-1B-Instruct (4M); Llama-3.2-3B-Instruct (1.1M); Llama-3.2-1B (905.7k). ## Other Licenses - [apache-2.0](https://savrn.com/models/licenses/apache-2-0) 3395 - [mit](https://savrn.com/models/licenses/mit) 1335 - [cc-by-nc-4.0](https://savrn.com/models/licenses/cc-by-nc-4-0) 215 - [cc-by-4.0](https://savrn.com/models/licenses/cc-by-4-0) 172 - [bsd-3-clause](https://savrn.com/models/licenses/bsd-3-clause) 108 - [gemma](https://savrn.com/models/licenses/gemma) 63 - [cc-by-nc-sa-4.0](https://savrn.com/models/licenses/cc-by-nc-sa-4-0) 56 - [unknown](https://savrn.com/models/licenses/unknown) 33 - [agpl-3.0](https://savrn.com/models/licenses/agpl-3-0) 32 - [llama3.1](https://savrn.com/models/licenses/llama3-1) 29 - [cc-by-sa-4.0](https://savrn.com/models/licenses/cc-by-sa-4-0) 20 - [gpl-3.0](https://savrn.com/models/licenses/gpl-3-0) 20 [See all](https://savrn.com/models/licenses)