# Text Generation Models, Page 9: Open-Weight AI Source: https://savrn.com/models/tasks/text-generation/page/9 Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- SAVRN Model Hub · Models by Task # Text Generation Models 1,582 open-weight text generation models in the SAVRN Model Hub, with Convergent Intelligence, Qwen and Francesca Padovani publishing the most. 1,582Models 732Publishers 107K to 1.7TParameter range 8Licenses ## SAVRN's Take Text generation is where most of the attention goes, and where most of the money is spent. The models on this page run from tiny research checkpoints that fit on a laptop to frontier-class systems that need a full server of accelerators. The first question we ask about any of them is not how smart it is, but what it costs to keep it answering around the clock. That cost comes down to three things: the size of the weights, the precision you run them at, and how much context each request carries. Moving from 16-bit to 4-bit cuts the memory a model needs by roughly four times, which is often the difference between one server and several. Mixture-of-experts models, such as the newer DeepSeek, Qwen and Nemotron releases, hold a large total parameter count but activate only a small share per token, so they answer faster than their size suggests while still needing the memory to hold every expert. Licensing decides the rest. Apache 2.0 and MIT models can go straight into a commercial product. Llama, Gemma's earlier generations and models marked other each carry terms you need to read first, and a model with no stated license is research material until the publisher says otherwise. Our advice is to shortlist by license, size to the hardware you can actually run, then test the survivors on your own prompts. The table below is sorted by what people download, not by what we recommend. SAVRN Research, 2026-09-18 ## Most Downloaded | Model | Publisher | Parameters | License | Monthly downloads | Cheapest GPUs at 16-bit | | --- | --- | --- | --- | --- | --- | | [Qwen3-0.6B](https://savrn.com/models/qwen3-0-6b) | Qwen | 752M | apache-2.0 | 25.2M | 1x MI300X, $1.85/hr | | [gpt2](https://savrn.com/models/gpt2) | OpenAI community | 137M | mit | 18.8M | 1x MI300X, $1.85/hr | | [Qwen3-4B](https://savrn.com/models/qwen3-4b) | Qwen | 4B | apache-2.0 | 11.4M | 1x MI300X, $1.85/hr | | [Qwen3-14B](https://savrn.com/models/qwen3-14b) | Qwen | 14.8B | apache-2.0 | 9.4M | 1x MI300X, $1.85/hr | | [Qwen3-8B](https://savrn.com/models/qwen3-8b) | Qwen | 8.2B | apache-2.0 | 8.1M | 1x MI300X, $1.85/hr | | [Qwen2.5-7B-Instruct](https://savrn.com/models/qwen2-5-7b-instruct) | Qwen | 7.6B | apache-2.0 | 6.4M | 1x MI300X, $1.85/hr | | [gpt-oss-20b](https://savrn.com/models/gpt-oss-20b) | OpenAI | 20.9B | apache-2.0 | 5.7M | 1x MI300X, $1.85/hr | | [opt-125m](https://savrn.com/models/opt-125m) | AI at Meta | — | other | 5.6M | — | | [DeepSeek-V3.2](https://savrn.com/models/deepseek-v3-2) | DeepSeek | 685.4B | mit | 5.4M | 7x MI325X, $14.00/hr | | [Qwen2.5-0.5B-Instruct](https://savrn.com/models/qwen2-5-0-5b-instruct) | Qwen | 494M | apache-2.0 | 5.3M | 1x MI300X, $1.85/hr | ## Licenses | License | Models | Commercial use | | --- | --- | --- | | [apache-2.0](https://savrn.com/models/licenses/apache-2-0) | 734 | Yes | | not stated | 345 | Not stated | | other | 175 | Read the license | | [mit](https://savrn.com/models/licenses/mit) | 152 | Yes | | [cc-by-nc-4.0](https://savrn.com/models/licenses/cc-by-nc-4-0) | 42 | Not without separate permission | | [llama3.2](https://savrn.com/models/licenses/llama3-2) | 33 | Read the license | | [gemma](https://savrn.com/models/licenses/gemma) | 22 | Yes, with conditions | | [llama3.1](https://savrn.com/models/licenses/llama3-1) | 21 | Yes, with conditions | ## Who Publishes Them | Publisher | Models | | --- | --- | | [Convergent Intelligence](https://savrn.com/model-publishers/reaperdoesntknow) | 54 | | [Qwen](https://savrn.com/model-publishers/qwen) | 42 | | [Francesca Padovani](https://savrn.com/model-publishers/francesca9805) | 37 | | [Yanggangu](https://savrn.com/model-publishers/yanggangu) | 28 | | [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) | 25 | | [Kumar](https://savrn.com/model-publishers/sandeep123) | 23 | ## All 1,582 Models, Page 9 of 27 Model · Text generation ### [olmo2-7b-biomed-chem](https://savrn.com/models/olmo2-7b-biomed-chem) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) Vigyan-7B-BioMed-Chem is a domain-specialized LoRA adapter trained on OLMo-2-1124-7B dedicated to organic chemical synthesis, pharmacology, and molecular biology. Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/olmo2-7b-biomed-chem) Model · Text generation ### [olmo2-7b-phd-pure-math](https://savrn.com/models/olmo2-7b-phd-pure-math) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) Vigyan-7B-PhD-Pure-Math is a domain-specialized LoRA adapter on OLMo-2-1124-7B optimized for rigorous axiomatic mathematics, abstract algebra, and topology. Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/olmo2-7b-phd-pure-math) Model · Text generation ### [GLM-5.3-ewaste-edition-GGUF](https://savrn.com/models/glm-5-3-ewaste-edition-gguf) [SixVolts](https://savrn.com/model-publishers/sixvolts) This is a quantization of GLM-5.3 that is optimized to work better on older hardware. This quant was built using i-matrix GGUF quantizations with the same principle as the GLM-5.2 e-waste edition: K-quants for the routed experts, not codebook i-quants, because on pre-AVX-512 CPUs and previous-gen datacenter GPUs (MI100 / gfx908) expert dequantization, not bandwidth, is the decode bottleneck. GLM-5.3 is GLM-5.2 with further post-training (identical architecture and config), so the 5.2 recipes carry over tensor for tensor. All are imatrix-quantized with unsloth's GLM-5.3 importance matrix (209 chunks × 3584 tokens). The per-tensor types are the 5.2 files' types read from their GGUF headers… Open weights mit [View model](https://savrn.com/models/glm-5-3-ewaste-edition-gguf) Model · Text generation ### [qwen0.5b-int4-Octuma](https://savrn.com/models/qwen0-5b-int4-octuma) [Omar Imanol Rodríguez Rodríguez](https://savrn.com/model-publishers/imanol11) Built with Qwen. Qwen/Qwen2.5-0.5B-Instruct cuantizado a INT4 con (1.91x mas chico) perdiendo 3.73% de calidad. Incluye las dos formas de usarlo: qwen0.5b-int4.gguf para llama.cpp y Android, y la carpeta.tq para PyTorch. Perplejidad en wikitext-2 test, 20 ventanas de 2048 tokens, todo medido dentro de llama.cpp con llama-perplexity para que la comparacion sea de tu a tu. En este tamaño, llama.cpp Q4KM rinde mejor (+2.63% contra +3.73%) y ocupa menos. La ventaja de Octuma aparece en modelos mas grandes: en Qwen2.5-3B la relacion se invierte. La tabla esta aqui para que se vea, no para esconderla. La comparativa de los tres tamaños esta en En PyTorch el modelo cuantizado genera mas lento que… Open weights apache-2.0 32,768 tokens gguf [View model](https://savrn.com/models/qwen0-5b-int4-octuma) Model · Text generation ### [cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-groot16-s2](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-groot16-s2) [Alexander Gurung](https://savrn.com/model-publishers/agurung) An OpenRLHF GRPO reinforcement-learning checkpoint for Qwen3-4B. - Saved at global step 88 of RL run seededrlbaseramp25stoppengen4kep2ncp5n3ncgroot16s2. - This is the best checkpoint by pass@8 so far in this run. Trained and validated on the cobalt-train ≤2/64 frontier (canonical cleaneval prompts): 1833 train / 112 held-out val problems the base model solved on at most 2 of 64 samples under the iidcanonical@64 hardness scan. Val evals sample at temperature 1.0 (matching the cleaneval frontier eval). Reward signal: binary code-correctness (1.0 if the generated program passes the problem's tests, otherwise 0.0). This checkpoint is the main revision (git branch) of the repo, with the model at… Open weights 4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-groot16-s2) Model · Text generation ### [Network-Cloud-Ops-Engineer-9B](https://savrn.com/models/network-cloud-ops-engineer-9b) [Reponx](https://savrn.com/model-publishers/reponx) Reponx 9B is a specialized AI model for network engineering, network operations, cloud networking, troubleshooting, and infrastructure operations. It is designed to help network and cloud engineers investigate technical issues using an evidence-first and safety-conscious troubleshooting approach. 1. Diagnose the likely problem. 2. Gather evidence before making changes. 3. Recommend read-only checks and commands first. 4. Explain what the results mean. 5. Suggest controlled remediation when appropriate. 6. Include validation and rollback considerations. Reponx 9B is a PEFT/LoRA fine-tuned model based on Qwen/Qwen3.5-9B. Load the Qwen3.5-9B base model with the Reponx adapter: Reponx provides… Open weights apache-2.0 peft [View model](https://savrn.com/models/network-cloud-ops-engineer-9b) Model · Text generation ### [Qwen3.8-27B-pi-mlx-6Bit](https://savrn.com/models/qwen3-8-27b-pi-mlx-6bit) [SolsticeAI](https://savrn.com/model-publishers/solsticeai) The Model SolsticeAI/Qwen3.8-27B-pi-mlx-6Bit was converted to MLX format from bytkim/Qwen3.8-27B-pi using mlx-lm version 0.31.2. Open weights apache-2.0 26.9B parameters 262,144 tokens transformers [View model](https://savrn.com/models/qwen3-8-27b-pi-mlx-6bit) Model · Text generation ### [Qwen3.6-35B-A3B-Distill-Qwen3.8-27B-GGUF](https://savrn.com/models/qwen3-6-35b-a3b-distill-qwen3-8-27b-gguf) [Lododo.org](https://savrn.com/model-publishers/lododoai) A distilled fine-tune of the Qwen3.6-35B-A3B base (MoE, 3B active), trained on data generated by Qwen3.8-27B as the teacher — bringing the base model to (and beyond) the teacher's capability while keeping the MoE speed advantage. - LoRA adapter (40 MB, mount on the base, saves 22 GB) — lododoai/Qwen3.6-35B-A3B-Distill-Qwen3.8-27B-GGUF-lora - Community work, not affiliated with or endorsed by Qwen. Base model: Qwen3.6-35B-A3B (Apache-2.0). moefast is a small launcher that configures llama.cpp for your machine. It reads the model's GGUF metadata, measures your free VRAM, computes the expert-cache size, context and batch parameters, writes them into an editable config file next to the model… Open weights apache-2.0 transformers [View model](https://savrn.com/models/qwen3-6-35b-a3b-distill-qwen3-8-27b-gguf) Model · Text generation ### [deepseek-v4-flash-news2026-lora](https://savrn.com/models/deepseek-v4-flash-news2026-lora) [Apoorv Saxena](https://savrn.com/model-publishers/apoorvumang) A LoRA adapter that teaches deepseek-ai/DeepSeek-V4-Flash-0731 world events from April 2025 to August 2026, past its knowledge cutoff, by continued pretraining on 909M tokens of real news — no synthetic data, no QA pairs. Code, evaluation harnesses and the full recipe: Most of the value in this repository is the findings section below, not the weights. This adapter works on prominent entities and fails on the long tail; the measurements explaining why are more useful than the checkpoint. Scored against 705 facts extracted from its own training corpus, each corroborated by ≥2 independent news domains with every source agreeing on the date: The facts were in the training data. 909M tokens of… Open weights other peft [View model](https://savrn.com/models/deepseek-v4-flash-news2026-lora) Model · Text generation ### [kas-4b](https://savrn.com/models/kas-4b) [Kashav Piya](https://savrn.com/model-publishers/kpiya) A fine-tuned decision engine based on Qwen/Qwen3-4B, submitted to the Decision Index v0.2.1 leaderboard. Decision Index: 40.1 (raw 54.18) — above Tev1-4B (29.2), within 6 points of Scion v4 9B (45.9). Per-area breakdown against all 4B submissions on Decision Index 0.2.1. kas-4b leads the 4B tier on: - Knowledge & Reasoning (39.5) — highest of all 4B models; higher LoRA rank (r64) likely helps on harder reasoning tasks - Arts & Human Taste (37.7) — highest of all 4B models by a wide margin - Language Understanding (35.2) — 13–25 points behind ezjev and Nox; language-heavy fine-tuning data favors those models - Tools & Automation (49.0) — behind ezjev (69.9) and Nox (60.1) Profile: the most… Open weights apache-2.0 4B parameters 40,960 tokens [View model](https://savrn.com/models/kas-4b) Model · Text generation ### [trade-idp-qwen3vl-model-v1](https://savrn.com/models/trade-idp-qwen3vl-model-v1) [Bisal Saha](https://savrn.com/model-publishers/bisalsaha) 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). - PEFT 0.20.0 Open weights peft [View model](https://savrn.com/models/trade-idp-qwen3vl-model-v1) 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 ### [Vigyan-7B-STEM-Instruct-v1](https://savrn.com/models/vigyan-7b-stem-instruct-v1) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) 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). - PEFT 0.19.1 Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/vigyan-7b-stem-instruct-v1) Model · Text generation ### [Cyber-F1-smoke](https://savrn.com/models/cyber-f1-smoke) [Autumn](https://savrn.com/model-publishers/autumn10) This model is a fine-tuned version of DuyTa/Cyber-F1. It has been trained using TRL. This model was trained with SFT. - PEFT 0.21.0 Open weights 35.1B parameters 262,144 tokens peft [View model](https://savrn.com/models/cyber-f1-smoke) Model · Text generation ### [cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-base-s3](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-base-s3) [Alexander Gurung](https://savrn.com/model-publishers/agurung) An OpenRLHF GRPO reinforcement-learning checkpoint for Qwen3-4B. - Saved at global step 66 of RL run seededrlbaseramp25stoppengen4kep2ncp5n3ncbases3. - This is the best checkpoint by pass@8 so far in this run. Trained and validated on the cobalt-train ≤2/64 frontier (canonical cleaneval prompts): 1833 train / 112 held-out val problems the base model solved on at most 2 of 64 samples under the iidcanonical@64 hardness scan. Val evals sample at temperature 1.0 (matching the cleaneval frontier eval). Reward signal: binary code-correctness (1.0 if the generated program passes the problem's tests, otherwise 0.0). This checkpoint is the main revision (git branch) of the repo, with the model at… Open weights 4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-base-s3) Model · Text generation ### [DeepSeek-R1-Distill-Qwen-1.5B-RKLLM](https://savrn.com/models/deepseek-r1-distill-qwen-1-5b-rkllm) [Hanzo Huang](https://savrn.com/model-publishers/hanzohuang) Text generation on Rockchip RK3576 and RK3588, converted from deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B. 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… Open weights mit rkllm [View model](https://savrn.com/models/deepseek-r1-distill-qwen-1-5b-rkllm) Model · Text generation ### [olmo2-7b-silicon-rtl-eda](https://savrn.com/models/olmo2-7b-silicon-rtl-eda) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) Vigyan-7B-Silicon-RTL-EDA is a domain-specialized Low-Rank Adaptation (LoRA) adapter built upon AllenAI's open-weights OLMo-2-1124-7B architecture. It is fine-tuned specifically for Electronic Design Automation (EDA), register-transfer level (RTL) Verilog synthesis, and formal verification assertions. Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/olmo2-7b-silicon-rtl-eda) Model · Text generation ### [StandardOne-3B-FP8](https://savrn.com/models/standardone-3b-fp8) [Standard Thinking](https://savrn.com/model-publishers/standardthinking) StandardOne-3B-FP8 is an FP8 (compressed-tensors, float8e4m3 weights, dynamic per-token activations) quantization of the released StandardOne-3B decision model. The language-model linear projections (q/k/v/o, gate/up/down) are quantized per-channel FP8 E4M3 with dynamic FP8 activations (llm-compressor's data-free FP8DYNAMIC recipe, no calibration data required); the vision tower, multi-modal projector, embeddings and lmhead are left unquantized in BF16. It was produced from source revision 68dafd17ead9b8cf6f85c4f08f7f2f3a1e7b9e5c of StandardOne-3B on 2026-09-25 using llm-compressor 0.14.0 (torch 2.14.0, transformers 5.17.0, compressed-tensors 0.19.0); results below. Served through SGLang… Open weights apache-2.0 3.8B parameters 262,144 tokens transformers [View model](https://savrn.com/models/standardone-3b-fp8) 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 ### [cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-iid16-s3](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-iid16-s3) [Alexander Gurung](https://savrn.com/model-publishers/agurung) An OpenRLHF GRPO reinforcement-learning checkpoint for Qwen3-4B. - Saved at global step 64 of RL run seededrlbaseramp25stoppengen4kep2ncp5n3nciid16s3. - This is the best checkpoint by pass@8 so far in this run. Trained and validated on the cobalt-train ≤2/64 frontier (canonical cleaneval prompts): 1833 train / 112 held-out val problems the base model solved on at most 2 of 64 samples under the iidcanonical@64 hardness scan. Val evals sample at temperature 1.0 (matching the cleaneval frontier eval). Reward signal: binary code-correctness (1.0 if the generated program passes the problem's tests, otherwise 0.0). This checkpoint is the main revision (git branch) of the repo, with the model at… Open weights 4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-iid16-s3) Model · Text generation ### [GradingQlora](https://savrn.com/models/gradingqlora) [SmuFypTeam5](https://savrn.com/model-publishers/smufypteam5) This model is a fine-tuned version of Qwen/Qwen3.5-9B. It has been trained using TRL. This model was trained with SFT. - PEFT 0.21.1 Open weights peft [View model](https://savrn.com/models/gradingqlora) Model · Text generation ### [StandardOne-8B-FP8](https://savrn.com/models/standardone-8b-fp8) [Standard Thinking](https://savrn.com/model-publishers/standardthinking) StandardOne-8B-FP8 is an FP8 (compressed-tensors, float8e4m3 weights, dynamic per-token activations) quantization of the released StandardOne-8B decision model. The language-model linear projections (q/k/v/o, gate/up/down) are quantized per-channel FP8 E4M3 with dynamic FP8 activations (llm-compressor's data-free FP8DYNAMIC recipe, no calibration data required); the vision tower, multi-modal projector, embeddings and lmhead are left unquantized in BF16. It was produced from source revision e88423700bb5ab9b2f50e176cf19825914345272 of StandardOne-8B on 2026-09-25 using llm-compressor 0.14.0 (torch 2.14.0, transformers 5.17.0, compressed-tensors 0.19.0); results below. Served through SGLang… Open weights apache-2.0 8.9B parameters 262,144 tokens transformers [View model](https://savrn.com/models/standardone-8b-fp8) Model · Text generation ### [flint-uncensored](https://savrn.com/models/flint-uncensored) [Ralfi Kier](https://savrn.com/model-publishers/electroglyph) Flint is Recursive’s mixture-of-experts language model for coding agents, with 39.8 billion total parameters and approximately 2.5 billion active per token. It is designed to investigate codebases, fix bugs, implement features, and write tests. Flint supports text input and output, reasoning, and multi-turn tool use. In our release evaluations, it scores 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2, and 74.2% on SWE-bench Verified. This repository contains the BF16 checkpoint, model configuration, tokenizer, chat template, and an SGLang serving kit. Quantized variants are prepared separately for NVIDIA and Apple Silicon runtimes. Flint results use temperature 1.0, top-p 0.95, and… Open weights apache-2.0 39.8B parameters 262,144 tokens transformers [View model](https://savrn.com/models/flint-uncensored) Model · Text generation ### [cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-groot16-s3](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-groot16-s3) [Alexander Gurung](https://savrn.com/model-publishers/agurung) An OpenRLHF GRPO reinforcement-learning checkpoint for Qwen3-4B. - Saved at global step 70 of RL run seededrlbaseramp25stoppengen4kep2ncp5n3ncgroot16s3. - This is the best checkpoint by pass@8 so far in this run. Trained and validated on the cobalt-train ≤2/64 frontier (canonical cleaneval prompts): 1833 train / 112 held-out val problems the base model solved on at most 2 of 64 samples under the iidcanonical@64 hardness scan. Val evals sample at temperature 1.0 (matching the cleaneval frontier eval). Reward signal: binary code-correctness (1.0 if the generated program passes the problem's tests, otherwise 0.0). This checkpoint is the main revision (git branch) of the repo, with the model at… Open weights 4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-groot16-s3) Model · Text generation ### [gemma-3-4b-it-RKLLM](https://savrn.com/models/gemma-3-4b-it-rkllm) [Hanzo Huang](https://savrn.com/model-publishers/hanzohuang) Text generation on Rockchip RK3576 and RK3588, converted from google/gemma-3-4b-it. The files currently published here are text-only. The source model supports images, but this conversion does not include a vision encoder. 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 w4a16 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… Open weights gemma rkllm [View model](https://savrn.com/models/gemma-3-4b-it-rkllm) Model · Text generation ### [olmo2-7b-astro-logic](https://savrn.com/models/olmo2-7b-astro-logic) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) Vigyan-7B-Astro-Logic is a domain-specialized LoRA adapter for OLMo-2-1124-7B fine-tuned for orbital mechanics, gravitational physics, and celestial dynamics. Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/olmo2-7b-astro-logic) Model · Text generation ### [alea-r4-3B-A0.7B-Base-ckpts](https://savrn.com/models/alea-r4-3b-a0-7b-base-ckpts) [ALEA Institute](https://savrn.com/model-publishers/alea-institute) Qwen3-MoE architecture; 3.32B parameters in total (3,315,557,632, tied input/output embeddings), 673M active per token. These are intermediate base-model pretraining checkpoints, not instruction-tuned. Safety and behaviour status: pending final checks. They are published for research: to follow what the model learns over training and to load any checkpoint with stock transformers. They are not a release. R4 is ALEA's next model after R3, trained only on lawful data: U.S. legal, regulatory and financial public-record text from the KL3M Data Project, plus synthetic, dictionary, knowledge and task data created by ALEA. main holds only this card. Each checkpoint is its own branch; load one with… Open weights cc-by-4.0 transformers [View model](https://savrn.com/models/alea-r4-3b-a0-7b-base-ckpts) Model · Text generation ### [cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-vs16-s3](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-vs16-s3) [Alexander Gurung](https://savrn.com/model-publishers/agurung) An OpenRLHF GRPO reinforcement-learning checkpoint for Qwen3-4B. - Saved at global step 60 of RL run seededrlbaseramp25stoppengen4kep2ncp5n3ncvs16s3. - This is the best checkpoint by pass@8 so far in this run. Trained and validated on the cobalt-train ≤2/64 frontier (canonical cleaneval prompts): 1833 train / 112 held-out val problems the base model solved on at most 2 of 64 samples under the iidcanonical@64 hardness scan. Val evals sample at temperature 1.0 (matching the cleaneval frontier eval). Reward signal: binary code-correctness (1.0 if the generated program passes the problem's tests, otherwise 0.0). This checkpoint is the main revision (git branch) of the repo, with the model at… Open weights 4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-vs16-s3) Model · Text generation ### [Qwen3.6-35B-A3B-heretic-lora](https://savrn.com/models/qwen3-6-35b-a3b-heretic-lora) [Alisher Amirbek](https://savrn.com/model-publishers/alisheramirbek) 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). - PEFT 0.19.1 Open weights peft [View model](https://savrn.com/models/qwen3-6-35b-a3b-heretic-lora) Model · Text generation ### [Kimibrain-Uncensored](https://savrn.com/models/kimibrain-uncensored) [Mohammed Ali Ali Al Marzooqi](https://savrn.com/model-publishers/burabeeh) Kimibrain is the open-weight, uncensored large language model released by Burabeeh Lab, founded by Mohammed Al Marzooqi — a National of the United Arab Emirates. This is the official GGUF build of Kimibrain, packaged by Burabeeh Lab for Ollama, llama.cpp, LM Studio and any other GGUF runtime. It ships with the optional vision projector, so the same file handles text and images. - Uncensored by design. Minimal refusal behaviour on legal creative, technical and research requests. Kimibrain answers the question instead of moralising about it. - Identity-locked. Kimibrain always identifies itself as Kimibrain, by Burabeeh Lab. The identity is carried by the chat template inside the GGUF, not… Open weights apache-2.0 llama.cpp [View model](https://savrn.com/models/kimibrain-uncensored) Model · Text generation ### [EdgeRock-4B-AWQ](https://savrn.com/models/edgerock-4b-awq) [Aaron Xu](https://savrn.com/model-publishers/hkust-aaron) AWQ W4A16 weights for Rock Matrix's EdgeRock-4B, with a packed INT4 CUDA runtime. Fixed memory for long-running generation. This EdgeRock-4B deployment package retains recurrent state, local attention, and Rock Memory Anchors (RMA), keeping persistent history at a fixed capacity as context grows. 固定记忆，长程生成。 本版本保留 EdgeRock-4B 的递推状态、局部注意力和记忆岩锚（RMA），历史内存不随上下文增长。 Rock Matrix Whitepaper / 岩芯矩阵技术白皮书： whitepaper connects EdgeRock's fixed memory, DiffusionRock's parallel generation, and SiliconRock's model-native silicon into the route toward Ultra-Fast Token. 白皮书讲清固定记忆、并行生成、模型固化三条路线，以及它们如何走向更快、更经济的生成。 EdgeRock combines recurrent state, sliding-window attention and Rock Memory Anchors (RMA) to… Open weights other 4.6B parameters 262,144 tokens transformers [View model](https://savrn.com/models/edgerock-4b-awq) Model · Text generation ### [ct0-alias3-micro-mlc](https://savrn.com/models/ct0-alias3-micro-mlc) [Eduardo Worrel](https://savrn.com/model-publishers/eduardoworrel) MLC-compiled weights of aliasrobotics/alias3-micro-1026 (CT0 — a 4B hybrid qwen35text: 24 Gated-DeltaNet linear-attention + 8 full-attention layers) for in-browser inference via WebLLM (WebGPU). This is the q4f161 build (~2.2 GB) — the recommended default. Higher-fidelity sibling: q4f321. The ONNX/ort-web path is capped by the browser's ~4 GB WASM limit (forcing an int4 embedding hack). WebLLM keeps weights in WebGPU buffers, so this build loads and runs with no 4 GB wall — standard generate(), higher fidelity than the int4 ONNX. tiewordembeddings + matching arch ⇒ the official prebuilt Qwen3.5-4B model-lib wasm is reusable (no compile): Manifest: the repo ships both ndarray-cache.json and… Open weights other mlc-llm [View model](https://savrn.com/models/ct0-alias3-micro-mlc) Model · Text generation ### [ProactiveInquirer-Qwen3-8B](https://savrn.com/models/proactiveinquirer-qwen3-8b) [Ido Levy](https://savrn.com/model-publishers/dolev31) Ido Levy 1,2 · 1,2 This is the trained questioner from Asking for What Was Never Requested: Horizontal and Vertical Proactivity in Agents. It is a LoRA adapter on Qwen3-8B, trained with Q&D (questioner and drafter). A tool-using agent usually does what it is asked, yet a task often needs information the user never mentions. This model decides, one step at a time, which question to send to the agent's retriever next, or that it is time to stop. It goes after two kinds of unstated need: ZIP code, so the agent looks up the account. the order names the product, and the product lists the size-8 variant. These numbers are from the paper, on held-out test splits. The MuSiQue reading compares… Open weights apache-2.0 peft [View model](https://savrn.com/models/proactiveinquirer-qwen3-8b) Model · Text generation ### [EdgeRock-4B-GPTQ](https://savrn.com/models/edgerock-4b-gptq) [Aaron Xu](https://savrn.com/model-publishers/hkust-aaron) GPTQ W4A16 weights for Rock Matrix's EdgeRock-4B, with a packed INT4 CUDA runtime. Fixed memory for long-running generation. This EdgeRock-4B deployment package retains recurrent state, local attention, and Rock Memory Anchors (RMA), keeping persistent history at a fixed capacity as context grows. 固定记忆，长程生成。 本版本保留 EdgeRock-4B 的递推状态、局部注意力和记忆岩锚（RMA），历史内存不随上下文增长。 Rock Matrix Whitepaper / 岩芯矩阵技术白皮书： whitepaper connects EdgeRock's fixed memory, DiffusionRock's parallel generation, and SiliconRock's model-native silicon into the route toward Ultra-Fast Token. 白皮书讲清固定记忆、并行生成、模型固化三条路线，以及它们如何走向更快、更经济的生成。 EdgeRock combines recurrent state, sliding-window attention and Rock Memory Anchors (RMA) to… Open weights other 4.6B parameters 262,144 tokens transformers [View model](https://savrn.com/models/edgerock-4b-gptq) Model · Text generation ### [StandardOne-3B-LoRA](https://savrn.com/models/standardone-3b-lora) [Standard Thinking](https://savrn.com/model-publishers/standardthinking) Standard One scores a bounded set of answers for a supplied scenario and returns probabilities through POST /v1/systemone. It does not generate free-form response text. This repository contains the 3B LoRA adapter and a merge recipe; serving requires a merged checkpoint and the server code. In the reported served evaluations, 3B has a lower median latency on the measured short-request profile; 8B scores higher on the public standard and hard tiers. See Benchmarks for the measurement conditions and limitations. The figure combines results from different measurement paths. See Benchmarks for served versus offline conditions; measured 24–26 September 2026. - Send a state and a bounded rubric… Open weights apache-2.0 peft [View model](https://savrn.com/models/standardone-3b-lora) Model · Text generation ### [EdgeRock-4B-NVFP4](https://savrn.com/models/edgerock-4b-nvfp4) [Aaron Xu](https://savrn.com/model-publishers/hkust-aaron) EdgeRock's constant-memory reasoning architecture with packed NVFP4 weights. Fixed memory for long-running generation. This EdgeRock-4B deployment package retains recurrent state, local attention, and Rock Memory Anchors (RMA), keeping persistent history at a fixed capacity as context grows. 固定记忆，长程生成。 本版本保留 EdgeRock-4B 的递推状态、局部注意力和记忆岩锚（RMA），历史内存不随上下文增长。 Rock Matrix Whitepaper / 岩芯矩阵技术白皮书： whitepaper connects EdgeRock's fixed memory, DiffusionRock's parallel generation, and SiliconRock's model-native silicon into the route toward Ultra-Fast Token. 白皮书讲清固定记忆、并行生成、模型固化三条路线，以及它们如何走向更快、更经济的生成。 This text checkpoint uses NVIDIA Model Optimizer's W4A16 NVFP4 format: E2M1 4-bit weights, one FP8… Open weights other 2.6B parameters 262,144 tokens transformers [View model](https://savrn.com/models/edgerock-4b-nvfp4) Model · Text generation ### [parallax-8b-lambda](https://savrn.com/models/parallax-8b-lambda) [Chutes](https://savrn.com/model-publishers/chutesai) These are research checkpoints from lambda, a training run of Parallax that is still in progress. Parallax is a mixture-of-experts language model trained on consumer GPUs spread over several countries. The hosts have no direct connections to each other and synchronize over the public internet. - Base model. It is pretrained on a mixture of web, document, math, code, knowledge-focused and book text. It is not instruction-tuned, chat-tuned or safety-tuned, and it will continue text rather than follow instructions. - Mid-training. Each export is a snapshot of a run that has not finished. Later exports are expected to differ, and quality between exports is not monotonic. - Research artifact. It… Open weights mit [View model](https://savrn.com/models/parallax-8b-lambda) Model · Text generation ### [qwen2.5-3b-tombench-qlora](https://savrn.com/models/qwen2-5-3b-tombench-qlora) [Hong ji hyeong](https://savrn.com/model-publishers/nextmina) Adapters for the paper "QLoRA Fine-Tuning for Theory-of-Mind Reasoning in a 3B-Parameter Small Language Model: A ToMBench-Based Cross-Benchmark Generalization Study" (Applied Sciences, under revision). Base model: unsloth/Qwen2.5-3B-Instruct-bnb-4bit. QLoRA, r = 16, α = 16, dropout 0. Use the pinned environment in the code repository (requirements.lock.txt) and greedy decoding with the system prompt in the repository's notebook. Loading the same weights with PeftModel.frompretrained changes about 1.4% of predictions (all near-ties); see Section 4.9 of the paper. Under cyclic permutation of the answer options (five seeds), the item-level adapters lower SocialIQa accuracy (−4.14 points); the… Open weights other peft [View model](https://savrn.com/models/qwen2-5-3b-tombench-qlora) Model · Text generation ### [SajiloFit-FitnessAndNutrition-Qwen2.5-Finetuned](https://savrn.com/models/sajilofit-fitnessandnutrition-qwen2-5-finetuned) [Bibek sapkota](https://savrn.com/model-publishers/bibek111) This repository contains the selected PEFT LoRA adapter for SajiloFit AI, an English-language fitness and general nutrition assistant built on top of Qwen/Qwen2.5-7B-Instruct. This is an adapter-only repository. It does not contain the complete Qwen2.5-7B model weights. To use it, load the Qwen base model first and then attach this adapter with PEFT. During QLoRA training, the base model is loaded in a quantized form to reduce GPU-memory use while the LoRA parameters are trained. The uploaded artifact is a LoRA adapter and can be attached to either a normal BF16/FP16 base model or a compatible 4-bit base-model load. The adapter is intended to improve Qwen's behaviour for SajiloFit tasks… Open weights apache-2.0 peft [View model](https://savrn.com/models/sajilofit-fitnessandnutrition-qwen2-5-finetuned) Model · Text generation ### [sat-tutor-qwen3-8b](https://savrn.com/models/sat-tutor-qwen3-8b) [Gopi Adusumilli](https://savrn.com/model-publishers/agk4444) by AGK FIRE INC A QLoRA fine-tune of Qwen/Qwen3-8B that tutors for the SAT (math + reading & writing). Trained on ~17k SAT-style examples (60% math, 40% reading & writing), one epoch, 2048-token training length, with full step-by-step solutions plus hint-style answers. 160 held-out questions (100 math, 60 reading & writing). Identical prompt and decoding for both: greedy, 400 new tokens, thinking off, same Answer: X format instruction. Strict scoring = unparseable counted wrong. Reports: evalbase.json, evaltuned.json. The tuned adapter also answers far more math questions (97/100 vs 71/100) — base adapter overfit toward math. More RW training data is recommended before the teacher is… Open weights apache-2.0 peft [View model](https://savrn.com/models/sat-tutor-qwen3-8b) Model · Text generation ### [qwen-metaphor-unke-camada-16-zero-um](https://savrn.com/models/qwen-metaphor-unke-camada-16-zero-um) [Luisa Muniz Stellet](https://savrn.com/model-publishers/luisastellet) 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). Open weights 596M parameters 40,960 tokens transformers [View model](https://savrn.com/models/qwen-metaphor-unke-camada-16-zero-um) Model · Text generation ### [StandardOne-8B-LoRA](https://savrn.com/models/standardone-8b-lora) [Standard Thinking](https://savrn.com/model-publishers/standardthinking) Standard One scores a bounded set of answers for a supplied scenario and returns probabilities through POST /v1/systemone. It does not generate free-form response text. This repository contains the 8B LoRA adapter and a merge recipe; serving requires a merged checkpoint and the server code. In the reported served evaluations, 8B scores higher than 3B on the public standard and hard tiers; 3B has a lower median latency on the measured short-request profile. See Benchmarks for the measurement conditions and limitations. The figure combines results from different measurement paths. See Benchmarks for served versus offline conditions; measured 24–26 September 2026. - Send a state and a bounded… Open weights apache-2.0 peft [View model](https://savrn.com/models/standardone-8b-lora) Model · Text generation ### [zolai-qwen2.5-3b-lora](https://savrn.com/models/zolai-qwen2-5-3b-lora) [Peter Pau Sian Lian](https://savrn.com/model-publishers/peterpausianlian) LoRA / QLoRA adapter fine-tuned for Tedim Zolai (Zo; ZVS 2018 orthography) generation and bilingual Zo↔English assistance. Part of the Zolai-AI project — a RAG-first bilingual AI toolkit to preserve and teach Tedim Zolai. Generated and curated Zolai text for this project follows Tedim ZVS 2018: - Prefer native-speaker review for dialect naturalness; Hakha forms must not be treated as Tedim ZVS. - Generate / continue Tedim Zolai text under ZVS 2018 norms - Assist Zo↔English drafting (dictionary / parallel-corpus style tasks) - Experimentation for low-resource Tibeto-Burman NLP research - Load as a PEFT adapter on Qwen/Qwen2.5-3B-Instruct for apps, tutors, or RAG-backed bots (e.g. zolai.space… Open weights unknown peft [View model](https://savrn.com/models/zolai-qwen2-5-3b-lora) Model · Text generation ### [Darwin-397B-JGOS](https://savrn.com/models/darwin-397b-jgos) [FINAL_Bench](https://savrn.com/model-publishers/final-bench) Darwin-397B-JGOS is the largest and highest-scoring member of the Darwin family. Built on Qwen 3.5 397B as the base, it transplants the FFN (expert) strengths of multiple high-performance models through the Darwin V9 platform, producing a 397B-parameter Mixture-of-Experts model with ~17B active parameters per token. It reaches 90.9 % on GPQA Diamond with pure greedy decoding (single sample) — surpassing Darwin-28B-REASON (89.39 %, achieved with the Darwin-DELPHI test-time engine) without using any test-time engine at all. This is the highest GPQA Diamond score in the Darwin family to date. Darwin is VIDRAFT's measuring-result-driven reasoning model family — approximately 20 official models… Open weights apache-2.0 403.4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/darwin-397b-jgos) Model · Text generation ### [Llama-3.3-70B-Instruct-CQ-4bit](https://savrn.com/models/llama-3-3-70b-instruct-cq-4bit) [ThousandML](https://savrn.com/model-publishers/thousandml) Llama 3.3 70B Instruct quantized to 4-bit with Coded Quantization (CQ). CQ is a calibration-free 4-bit quantization method designed to preserve the behavior of the original model without requiring a calibration dataset. Because CQ does not optimize the quantization around a fixed calibration corpus, it is designed to generalize beyond calibration-specific data and workloads. This repository contains the 70B-parameter Llama 3.3 Instruct model quantized with Coded Quantization (CQ) 4-bit. Calibration-free. CQ does not require a representative calibration dataset before quantization. This avoids tying the quantization procedure to a particular calibration corpus and is intended to provide… Open weights llama3.3 4.2B parameters 131,072 tokens [View model](https://savrn.com/models/llama-3-3-70b-instruct-cq-4bit) Model · Text generation ### [qwen1.5b-int4-Octuma](https://savrn.com/models/qwen1-5b-int4-octuma) [Omar Imanol Rodríguez Rodríguez](https://savrn.com/model-publishers/imanol11) Built with Qwen. Qwen/Qwen2.5-1.5B-Instruct cuantizado a INT4 con (2.34x mas chico) perdiendo 2.11% de calidad. Incluye las dos formas de usarlo: qwen1.5b-int4.gguf para llama.cpp y Android, y la carpeta.tq para PyTorch. Perplejidad en wikitext-2 test, 20 ventanas de 2048 tokens, todo medido dentro de llama.cpp con llama-perplexity para que la comparacion sea de tu a tu. Este modelo pierde 2.2 veces menos calidad que llama.cpp Q4KM, el formato mas usado para correr modelos en local, a cambio de 0.34 GB mas de archivo. La comparativa de los tres tamaños esta en En PyTorch el modelo cuantizado genera mas lento que el original: Octuma desempaqueta los 4 bits en cada multiplicacion sin un… Open weights apache-2.0 32,768 tokens gguf [View model](https://savrn.com/models/qwen1-5b-int4-octuma) Model · Text generation ### [KeyVoice-Refiner-2B-v0.1](https://savrn.com/models/keyvoice-refiner-2b-v0-1) [岡優志 / Okayuji](https://savrn.com/model-publishers/okayuji) A text-refinement model for cleaning up speech-to-text output: punctuation, polite-form conversion, translation, summarization, and filler-word removal. This repository contains a complete Q4KM GGUF text model, ready for a compatible runtime. For: developers adding local text cleanup, or Mac users who want to try it in KeyVoice. Not for: general chat, code generation, mathematical reasoning, or long-form creative writing. Platform: GGUF runtimes with Qwen3.5 support; the original performance report used an Apple Silicon M4 Mac. Use KeyVoice for Mac for the app experience. For your own integration, download the file with the Hugging Face CLI: Run it with llama.cpp using its CLI instructions.… Open weights apache-2.0 gguf [View model](https://savrn.com/models/keyvoice-refiner-2b-v0-1) Model · Text generation ### [Shreyansh-STEM-AI-2B-v3-Distilled](https://savrn.com/models/shreyansh-stem-ai-2b-v3-distilled) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) 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). - PEFT 0.21.0 Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/shreyansh-stem-ai-2b-v3-distilled) Model · Text generation ### [cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-base-s2](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-base-s2) [Alexander Gurung](https://savrn.com/model-publishers/agurung) An OpenRLHF GRPO reinforcement-learning checkpoint for Qwen3-4B. - Saved at global step 64 of RL run seededrlbaseramp25stoppengen4kep2ncp5n3ncbases2. - This is the best checkpoint by pass@8 so far in this run. Trained and validated on the cobalt-train ≤2/64 frontier (canonical cleaneval prompts): 1833 train / 112 held-out val problems the base model solved on at most 2 of 64 samples under the iidcanonical@64 hardness scan. Val evals sample at temperature 1.0 (matching the cleaneval frontier eval). Reward signal: binary code-correctness (1.0 if the generated program passes the problem's tests, otherwise 0.0). This checkpoint is the main revision (git branch) of the repo, with the model at… Open weights 4B parameters 262,144 tokens transformers [View model](https://savrn.com/models/cobalt-seeded-rl-base-ramp25-stoppen-gen4k-ep2-ncp5-n3nc-base-s2) Model · Text generation ### [codeparrot-ds](https://savrn.com/models/codeparrot-ds) [Ai Xin](https://savrn.com/model-publishers/aixin2024) This model is a fine-tuned version of distilbert/distilgpt2 on an unknown dataset. The following hyperparameters were used during training: - learningrate: 0.0005 - trainbatchsize: 32 - evalbatchsize: 32 - gradientaccumulationsteps: 8 - totaltrainbatchsize: 256 - lrschedulertype: cosine - lrschedulerwarmupsteps: 1000 - numepochs: 1 - mixedprecisiontraining: Native AMP - Transformers 4.57.1 - Pytorch 2.8.0+cu126 - Datasets 4.0.0 - Tokenizers 0.22.1 Open weights apache-2.0 82M parameters transformers [View model](https://savrn.com/models/codeparrot-ds) Model · Text generation ### [exaone-7b-verireason-reproduced-1513-fullft-best-grpo-reproduced-1.0](https://savrn.com/models/exaone-7b-verireason-reproduced-1513-fullft-best-grpo-reproduced-1-0) [Jongbin Won](https://savrn.com/model-publishers/jongbin-kr) This model is a fine-tuned version of Jongbin-kr/exaone-7b-verireason-reproduced-1513-fullft-epoch4. It has been trained using TRL. This model was trained with GRPO, a method introduced in DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models. Open weights 7.8B parameters 32,768 tokens transformers [View model](https://savrn.com/models/exaone-7b-verireason-reproduced-1513-fullft-best-grpo-reproduced-1-0) Model · Text generation ### [DeepSeek-R1-7B-MathPhysics-v7-lora](https://savrn.com/models/deepseek-r1-7b-mathphysics-v7-lora) [Luis Orlando Ponce](https://savrn.com/model-publishers/astorlui77) 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). - PEFT 0.18.1 Open weights peft [View model](https://savrn.com/models/deepseek-r1-7b-mathphysics-v7-lora) Model · Text generation ### [thai-qa-lab-model](https://savrn.com/models/thai-qa-lab-model) [Punyapat](https://savrn.com/model-publishers/revespera) โมเดล Thai GPT-2 ที่ได้รับการ Fine-tune โดยนักศึกษา (revespera) สำหรับงานถาม-ตอบภาษาไทยจากชุดข้อมูลโรค 3,000 คู่ 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). Open weights mit 124M parameters [View model](https://savrn.com/models/thai-qa-lab-model) Model · Text generation ### [vigyan-1.5b-adapter-engineering](https://savrn.com/models/vigyan-1-5b-adapter-engineering) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) This repository contains an experimental domain-specialized LoRA adapter trained as part of the Vigyan AI Sovereign Mixture of Experts (MoE) Upcycling Initiative. The goal of this experimental line was to train isolated, high-rank domain experts on specialized STEM corpora and examine whether post-hoc upcycling via uncalibrated linear routers (mergekit-moe) could synthesize dense reasoning models into edge-deployable sparse MoE networks. During post-training MoE upcycling experiments, this adapter was used in a 4-expert sparse mixture configuration. The experimental run revealed pivotal architectural insights: 1. Attention-FFN Desynchronization: Because this LoRA was trained across both… Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/vigyan-1-5b-adapter-engineering) Model · Text generation ### [vigyan-1.5b-adapter-mathematics](https://savrn.com/models/vigyan-1-5b-adapter-mathematics) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) This repository contains an experimental domain-specialized LoRA adapter trained as part of the Vigyan AI Sovereign Mixture of Experts (MoE) Upcycling Initiative. The goal of this experimental line was to train isolated, high-rank domain experts on specialized STEM corpora and examine whether post-hoc upcycling via uncalibrated linear routers (mergekit-moe) could synthesize dense reasoning models into edge-deployable sparse MoE networks. During post-training MoE upcycling experiments, this adapter was used in a 4-expert sparse mixture configuration. The experimental run revealed pivotal architectural insights: 1. Attention-FFN Desynchronization: Because this LoRA was trained across both… Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/vigyan-1-5b-adapter-mathematics) Model · Text generation ### [vigyan-1.5b-adapter-technology](https://savrn.com/models/vigyan-1-5b-adapter-technology) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) This repository contains an experimental domain-specialized LoRA adapter trained as part of the Vigyan AI Sovereign Mixture of Experts (MoE) Upcycling Initiative. The goal of this experimental line was to train isolated, high-rank domain experts on specialized STEM corpora and examine whether post-hoc upcycling via uncalibrated linear routers (mergekit-moe) could synthesize dense reasoning models into edge-deployable sparse MoE networks. During post-training MoE upcycling experiments, this adapter was used in a 4-expert sparse mixture configuration. The experimental run revealed pivotal architectural insights: 1. Attention-FFN Desynchronization: Because this LoRA was trained across both… Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/vigyan-1-5b-adapter-technology) Model · Text generation ### [laya-heart-gemma-e2b-v4-GGUF](https://savrn.com/models/laya-heart-gemma-e2b-v4-gguf) [Yunkun Song](https://savrn.com/model-publishers/yunkunsong) A fine-tuned Gemma 4 E2B for Laya-Heart, a local web app that helps you reply in a couple's chat. Given the chat and Laya's reading of it, the model writes two candidate replies, one per line. You choose, edit and send one yourself. 297 fictional practice-exam chats from the project's train-kit (253 Chinese, 44 English). Both models drafted every chat the way the app does. The judge, Qwen3-32B-AWQ, did not know which was which and judged every pair twice with the sides swapped; ties count as half a win. The flags come from the app's own checks. They are rule-based, not reviewed by a person. Self-check: the trainer's self-check (50 chats, nine gates) passed eight gates and missed one: judge… Open weights apache-2.0 [View model](https://savrn.com/models/laya-heart-gemma-e2b-v4-gguf) Model · Text generation ### [vigyan-1.5b-adapter-science](https://savrn.com/models/vigyan-1-5b-adapter-science) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) This repository contains an experimental domain-specialized LoRA adapter trained as part of the Vigyan AI Sovereign Mixture of Experts (MoE) Upcycling Initiative. The goal of this experimental line was to train isolated, high-rank domain experts on specialized STEM corpora and examine whether post-hoc upcycling via uncalibrated linear routers (mergekit-moe) could synthesize dense reasoning models into edge-deployable sparse MoE networks. During post-training MoE upcycling experiments, this adapter was used in a 4-expert sparse mixture configuration. The experimental run revealed pivotal architectural insights: 1. Attention-FFN Desynchronization: Because this LoRA was trained across both… Open weights cc-by-nc-4.0 peft [View model](https://savrn.com/models/vigyan-1-5b-adapter-science) Model · Text generation ### [Vigyan-AI-32B-Titan-v1](https://savrn.com/models/vigyan-ai-32b-titan-v1) [Shreyansh singh](https://savrn.com/model-publishers/shreyansh12183) Vigyan-32B Titan is a heavyweight sovereign model designed for graduate-level scientific reasoning, semiconductor VLSI timing closure, aerospace astrodynamics, and formal mathematical proofs. Open weights cc-by-nc-4.0 [View model](https://savrn.com/models/vigyan-ai-32b-titan-v1) Model · Text generation ### [qwen3-8b-followup-lora](https://savrn.com/models/qwen3-8b-followup-lora) [Dev89125](https://savrn.com/model-publishers/chenhaodev) 用于健康自诊 / 电话分诊场景的追问模型。给定患者自诊的多轮对话上下文，在患者每轮最新回复后，生成 1-3 个最可能、最有鉴别价值的开放式追问问题，用于缩小诊断范围、评估风险（危险信号/red flag），并避免重复询问患者已告知的信息。 - 基础模型：Qwen/Qwen3-8B (使用 unsloth 4-bit 量化微调) - 训练方式：LoRA (r=16, rslora, QLoRA 4bit) + SFT，约 2 epochs，约 3016 steps - 多轮对话输入：接受任意长度的 患者(user)/猜想官(assistant) 累积多轮对话，输出下一轮追问。 - 跨领域整合：含跨知识单元（multi-hop）样本，支持在多系统/多病史叠加场景下综合追问。 共 约 99.5k 条多轮健康自诊追问样本（train 96.5k / eval 3k），由三个子集混合： 知识来源（用于 teacher 蒸馏，不直接发布其原始文本）： - 西氏内科学精要（中文，task2） - UpToDate 临床主题与患者教育内容（英文，task3，老师模型将其蒸馏成中文自诊对话） 数据由本地教师大模型（vLLM minimax）根据上述医学知识文本、按统一 schema 蒸馏生成，覆盖 起病/诱因、部位、性质、持续时间、加重缓解因素、伴随症状、危险信号、既往史、用药史、年龄性别、家族史 等多样鉴别维度。 - multi-hop 跨单元样本：要求综合两个相关领域的知识，避免模型只盯住某一窄知识点、抛掉其他信息。 - RAG 知识库（FAISS）：rag/… Open weights apache-2.0 peft [View model](https://savrn.com/models/qwen3-8b-followup-lora) ## Questions ### Which Text generation models are most downloaded? By monthly downloads reported by the Hugging Face Hub: olmo2-7b-biomed-chem (52); olmo2-7b-phd-pure-math (52); GLM-5.3-ewaste-edition-GGUF (51). ## Other Tasks - [Image and Text to Text](https://savrn.com/models/tasks/image-and-text-to-text) 588 - [Robotics](https://savrn.com/models/tasks/robotics) 403 - [Feature Extraction](https://savrn.com/models/tasks/feature-extraction) 400 - [Text Classification](https://savrn.com/models/tasks/text-classification) 335 - [Text to Image](https://savrn.com/models/tasks/text-to-image) 236 - [Speech Recognition](https://savrn.com/models/tasks/speech-recognition) 190 - [Text to Speech](https://savrn.com/models/tasks/text-to-speech) 148 - [Image Classification](https://savrn.com/models/tasks/image-classification) 140 - [Audio Classification](https://savrn.com/models/tasks/audio-classification) 130 - [Sentence Similarity](https://savrn.com/models/tasks/sentence-similarity) 103 - [Reinforcement Learning](https://savrn.com/models/tasks/reinforcement-learning) 102 - [Token Classification](https://savrn.com/models/tasks/token-classification) 101 [See all](https://savrn.com/models/tasks)