weighted/imatrix quants of https://huggingface.co/z51722369/ZOZ-Function-Master-3B For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/ZOZ-Function-Master-3B-GGUF If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9…
SAVRN Model Hub
Open-Weight Models
An open-weight model is an AI model whose trained weights are published for anyone to download. The weights are what the model learned in training. With a copy of them you can run the model on hardware you control and train it further on your own data.
Open weights are not the same as open source. Many publishers release the weights without the training data or code, and the license sets what you may do with the model. This library puts each model's full card, architecture, files, license and published evaluations on one page.
12,283Models
3,510Datasets
290Papers
7,436Publishers
12,632Sourced relationships
Updated 2026-10-22 · How the library is built
12,283 models, sorted by most downloaded.
This is the model card of a transformers model that has been pushed on the Hub. Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Use the code below to get started with the model. Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).
Testing You should use Zelretch to trigger the image generation. Download them in the Files & versions tab.
Testing You should use Zelretch22 to trigger the image generation. You should use old man to trigger the image generation. You should use zelretchclothes to trigger the image generation. Download them in the Files & versions tab.
A ~565M-parameter language model that interleaves Mamba-3 (MIMO) and Gated DeltaNet-2 (GDN-2) sequence mixers with two MCEN stages (a reversible fixed-point "reasoning" core with a delta-rule memory), trained from scratch on the Zenyx-V3 pretraining mixture with the DeepSeek tokenizer. This repository holds training checkpoints, logs and diagnostics of an ongoing run; it is not a finished release. Data mixture — 121 sources in the frozen manifest (data/manifestv2.json, created 2026-10-10 06:04:26, hash 9454e840c2b9f032…); grouped shares: The comparison models are the same author's: Zenyx-v2-base (85M unique parameters, 32 effective layers, 32k vocabulary) and Zenyx-V3-Base (1.58B-parameter…
Zenyx V3 is an efficient 1.5B-parameter Mixture-of-Experts (MoE) foundation model built for low-latency inference and high throughput. It is written from scratch in JAX/Flax and trained on TPU v5e-8. active per token, with a Sinkhorn transport-based gate. latent subspace, cutting HBM bandwidth and memory footprint. stability at scale. YaRN and RoPE scaling factors are precomputed so context can be extended at inference time beyond the trained length. All tasks are evaluated with the standard base-model protocol: the model scores the log-likelihood of every candidate continuation and the highest-scoring one is taken as the answer. Nothing is generated and no output parsing is involved, so…
This repo is a post-training of Zero-WAM (zero-wam-pretrain, 10.8B) on 49 RoboCasa atomic tasks. The only task signal is a prompt video: the model never sees language, in training or at eval. The prompt is a video of the same task, taken from a different episode and a different scene. Everything in this repo was produced in a single run. All files are uncompressed (stored) zip archives. Full write-ups are inside results.zip: - report.md: overview of the experiment and its results. - reportpipeline.md: detailed record of how the data-conversion and training pipelines were built. Online, randomized RoboCasa target-split scenes, with prompt videos generated live by MimicGen. 10 scenarios per…
ZhSQL-R1-1.5B is a bilingual Text-to-SQL model based on Qwen2.5-Coder-1.5B, trained using supervised fine-tuning (SFT) followed by execution-reward GRPO. The model was fine-tuned for two epochs on 7,127 examples derived from the BIRD benchmark, with approximately 48% Chinese prompts. Key settings: LoRA rank 32, maximum sequence length 3,072, effective batch size 16, and learning rate 2e-4. Loss was applied only to the SQL response. The SFT model generated eight SQL candidates per question. Examples with partial success (0 < passrate < 1) were retained, producing 3,692 learnable examples. GRPO used binary execution-equivalence rewards for 300 steps, with eight generations per prompt, LoRA…
面向 RTX 5070 Ti 16GB 的单卡 looped language model 研究项目。当前模型有 803,629,472 个物理参数，可简称 800M 或 0.8B；默认 4 loops 时为 2 prefix + 8×4 shared body + 2 suffix，即 36 次逻辑 block application。 第一阶段只训练 CCI3-HQ 中文与 FineWeb-Edu 英文 Base Model。代码、Agent、长文档 和长上下文能力放在后续 continued pretraining / SFT，不混入 Stage A。 DATAROOT 可以放在项目外。程序递归发现所有子孙目录中的.jsonl，一级目录名 就是 source，默认要求如下： 每行至少包含字符串字段 text；若存在 id 或 documentid，它用于稳定文档划分， 否则使用正文 SHA-256。字段名、语言和验证比例在 configs/data-stage-a.json 配置。 训练不会生成或依赖固定 token.bin。首次扫描会建立一个轻量.npz/JSON 索引， 只记录文件偏移、token 数、source、language、document hash 和 train/validation split； 正文始终在训练时从原 JSONL 动态读取并 tokenize。原文件路径、大小、mtime、tokenizer 或 source 配置变化会产生新索引签名。 在 PowerShell 中运行： 它只创建项目内.venv，并按…
Zircon v2 is a 0.6B-parameter decision model from Fahrenheit Research. It runs fully on-device on Apple silicon. You give it an email, a message or a pending task plus a set of options, and it returns a calibrated probability for each option in under 50 ms per decision. (1) MacBook Pro (Apple M5), 8-bit weights, median. Speed varies by hardware. (2) Fahrenheit Research internal testing, September 2026, on held-out emails not seen in training. (3) Fahrenheit Research internal testing, September 2026. 400 cases (2,000 decisions) from the public LocalLLaMA typed-decisions test set. (4) Fahrenheit Research internal testing, September 2026, on game seeds not seen in training. Built on Qwen3-0.6B…
High-efficiency, hardware-tested GGUF releases of Qwen3.8-27B (27 Billion Parameters, Dense Architecture). All models in this repository have been physically converted, verified on hardware (AMD Instinct MI300X with ROCm / HIP), and benchmarked for prompt throughput, token generation velocity, and benchmark accuracy against baseline models. This repository provides three specialized model tiers: - zraldv1-ac (Accuracy-Priority Tier): 17.08 GiB (18.3 GB). Near-lossless retention (99.68% accuracy), matches or outperforms standard Q6K and Q80 quality while saving ~10 GB VRAM compared to Q80. - zraldv1-ba (Balanced Sweet-Spot): 14.46 GiB (15.5 GB). Optimal balance (99.12% accuracy), fits…
Official high-efficiency GGUF release of Qwen3.8-27B optimized for Accuracy Priority Tier. This repository contains zraldv1-ac.gguf (17.08 GiB / 18.34 GB), physically benchmarked on AMD Instinct MI300X hardware. For cross-comparison tables against standard Q80, Q6K, Q5K, Q4K, and Q2K models, visit the master repository: - Base model by the Qwen Team (Alibaba) under Apache 2.0. - Runtime by Georgi Gerganov and the llama.cpp community.
Official high-efficiency GGUF release of Qwen3.8-27B optimized for Balanced Sweet-Spot Tier. This repository contains zraldv1-ba.gguf (14.46 GiB / 15.52 GB), physically benchmarked on AMD Instinct MI300X hardware. For cross-comparison tables against standard Q80, Q6K, Q5K, Q4K, and Q2K models, visit the master repository: - Base model by the Qwen Team (Alibaba) under Apache 2.0. - Runtime by Georgi Gerganov and the llama.cpp community.
Official high-efficiency GGUF release of Qwen3.8-27B optimized for Compressed Size Tier. This repository contains zraldv1-cs.gguf (10.18 GiB / 10.93 GB), physically benchmarked on AMD Instinct MI300X hardware. For cross-comparison tables against standard Q80, Q6K, Q5K, Q4K, and Q2K models, visit the master repository: - Base model by the Qwen Team (Alibaba) under Apache 2.0. - Runtime by Georgi Gerganov and the llama.cpp community.
This is a trained model of a A2C agent playing PandaReachDense-v3 using the stable-baselines3 library.
This is a trained model of a A2C agent playing PandaReachDense-v3 using the stable-baselines3 library.
Model Collections
Hand-picked starting points, each with the reason it exists.
Collection · 4 entries
Embedding models for retrieval
Sentence and document embedding models used to build retrieval systems. Dimension and sequence length matter more than size here, and both come from the publisher.
Collection · 4 entries
Models that fit on one accelerator
Models whose publisher-reported parameter count puts them within reach of a single accelerator at common precisions. Memory needed depends on precision and serving configuration, so treat the parameter count as the starting point, not the answer.
Collection · 6 entries
Open-weight text models worth knowing
Widely used open-weight language models, chosen because each one is a distinct family rather than a variant of the one above it. Selection, not a ranking.
Collection · 3 entries
Speech and audio models
Recognition and synthesis models, grouped so the two directions are easy to compare.
Open-Weight Models Explained
What is an open-weight model?
An AI model whose trained weights are published for anyone to download, so it can be run, tested and fine-tuned on hardware the user controls.
Is an open-weight model the same as open source?
Not always. Open weights means the trained model can be downloaded. Open source usually also means the training code and data are available and the license allows broad reuse. Many open-weight models release the weights only.
Can I use an open-weight model commercially?
It depends on the license. Apache 2.0 and MIT allow commercial use. Other licenses limit it, for example to non-commercial use or below a set number of users. Every model page here shows its license.
How much memory does an open-weight model need?
About two bytes per parameter at 16-bit precision, so a 7-billion-parameter model needs roughly 14 GB for its weights, plus memory for the context it processes. Each model page lists its parameter count and the size of its files.
Related SAVRN Research
The hub sits beside SAVRN's market data and infrastructure research: what models cost to run, and what it takes to run them.
SAVRN Index
What open models cost to run
The same open-weight model priced by every host that serves it, per million tokens.
Research Hub
Data center trackers and maps
Moratoriums, permits, power, water and capital behind the facilities that run these models.
Method
How the Model Hub is built
Sources, evidence labels, refresh behaviour, and the limits of every comparison here.