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).
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.
Most Downloaded
| Model | Publisher | Parameters | License | Monthly downloads | Cheapest GPUs at 16-bit |
|---|---|---|---|---|---|
| Qwen3-0.6B | Qwen | 752M | apache-2.0 | 25.2M | 1x MI300X, $1.85/hr |
| gpt2 | OpenAI community | 137M | mit | 18.8M | 1x MI300X, $1.85/hr |
| Qwen3-4B | Qwen | 4B | apache-2.0 | 11.4M | 1x MI300X, $1.85/hr |
| Qwen3-14B | Qwen | 14.8B | apache-2.0 | 9.4M | 1x MI300X, $1.85/hr |
| Qwen3-8B | Qwen | 8.2B | apache-2.0 | 8.1M | 1x MI300X, $1.85/hr |
| Qwen2.5-7B-Instruct | Qwen | 7.6B | apache-2.0 | 6.4M | 1x MI300X, $1.85/hr |
| gpt-oss-20b | OpenAI | 20.9B | apache-2.0 | 5.7M | 1x MI300X, $1.85/hr |
| opt-125m | AI at Meta | — | other | 5.6M | — |
| DeepSeek-V3.2 | DeepSeek | 685.4B | mit | 5.4M | 7x MI325X, $14.00/hr |
| 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 | 734 | Yes |
| not stated | 345 | Not stated |
| other | 175 | Read the license |
| mit | 152 | Yes |
| cc-by-nc-4.0 | 42 | Not without separate permission |
| llama3.2 | 33 | Read the license |
| gemma | 22 | Yes, with conditions |
| llama3.1 | 21 | Yes, with conditions |
Who Publishes Them
| Publisher | Models |
|---|---|
| Convergent Intelligence | 54 |
| Qwen | 42 |
| Francesca Padovani | 37 |
| Yanggangu | 28 |
| Shreyansh singh | 25 |
| Kumar | 23 |
All 1,582 Models, Page 11 of 27
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).
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).
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).
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).
Upload in progress (started October 1, 2026). The 30 full-parameter checkpoints are being published sequentially. A checkpoint is downloadable once its directory appears; please wait for checkpoint-300 before treating the full set as complete. The run initialized from the original Qwen3-8B weights, not from an earlier SFT or RL checkpoint. It completed 300 optimizer steps using AgentEvolver. Checkpoints were saved every 10 steps, from checkpoint-10 through checkpoint-300. Each directory is a standalone Transformers model with weights, configuration, chat template, and tokenizer. The weights are FP32 and approximately 32.8 GB per checkpoint. Optimizer state is not included in this…
This repository releases a point-in-time, model-only export of the 49,097,218-parameter detachable causal-language-model student produced alongside the continuing 1,221,580,802-parameter AGILLM parent. The initial weight release does not contain the 1.22B parent weights or parent optimizer state. The released weight artifact was frozen from the student state associated with parent checkpoint step 2,560,770. The current single-file trainer source is now published below. Adaptive General Intelligence Large Language Model v4.4 is the release boundary for AGILLM's checkpoint-compatible, Matryoshka-inspired teacher-student system: - a mature 1,221,580,802-parameter parent that continues its…
A Qwen/Qwen3-8B policy trained as a multi-turn search agent (Search-R1 style) with Process-GRPO: a process reward model (Olmo-3-7B-Think verifier) scores each turn, with per-(group, turn-position) advantage normalization and verifier prompts that include the retrieved tool responses and the gold answer. The model at the repository root is the final policy (training step 320). Intermediate checkpoints are provided under step / subfolders. - step20/ — intermediate checkpoint at training step 20 - step40/ — intermediate checkpoint at training step 40 - step60/ — intermediate checkpoint at training step 60 - step80/ — intermediate checkpoint at training step 80 - step100/ — intermediate…
Local hardware: Mechrevo Kuangshi GM7AG0M — RTX 3060 Laptop 6GB GDDR6, 64GB DDR5, i7-12700H (14C/20T, 4.7GHz), Windows 11, Samsung 990 Pro. Good for imatrix and 0.6–35B-class work in RAM. 9B+ and searches need rented H200/Blackwell, typically $100 per quant. Released as part of the NOESIS Professional Multilingual Dubbing Automation Platform (framework: Deterministic Hybrid Control Framework for Frozen Neural Operators — DHCF-FNO). Repack (BF16 safetensors, shards ≤50 GB) and community GGUF quantizations of (modeltype: aliceai, 80B total / 3B active MoE with KDA layers). Исходные 49 шардов Yandex слиты в один стриминговый файл и заново нарезаны SHA256 каждого тензора, 0 расхождений (см.…
A general-purpose chat model on Qwen3-4B that knows when to call a web-search tool, and when not to. Part of the Amethyst family. Amethyst 2 Mini (Failed) keeps a direct, conversational voice and adds reliable web-search tool calling: search when the answer depends on current information, answer from knowledge when it does not, decompose multi-part questions into several queries, and summarize retrieved snippets into a cited answer. Do not disable thinking (enablethinking=False, --reasoning off, or a hand-built prompt that ends in an empty block). The training data taught tool calls in turns that have no empty think block, so with one present the model tends to answer in prose instead of…
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).
Block 0 receives the backbone's final normalized output. Blocks 1 and 2 apply their own pre-normalization. A shared, embedding-tied LM head supplies all three entropy evaluations and the final output prediction. A complete recurrent backbone, refined by three entropy-gated attention blocks. For a smaller download, see the 180M and 440M models below. AMOR (Adaptive Metacognitive Output Router) uses the model's predictive uncertainty to decide where attention should refine its recurrent representation. Each appended block reads normalized output entropy, compares it with a frozen threshold at inference, and adds an attention update only where its gate fires. The recurrent backbone and…
Block 0 receives the backbone's final normalized output. Blocks 1 and 2 apply their own pre-normalization. A shared, embedding-tied LM head supplies all three entropy evaluations and the final output prediction. A complete recurrent backbone, refined by three entropy-gated attention blocks. Looking for the larger model? AMOR-Gated DeltaNet 1.5B. AMOR (Adaptive Metacognitive Output Router) uses the model's predictive uncertainty to decide where attention should refine its recurrent representation. Each appended block reads normalized output entropy, compares it with a frozen threshold at inference, and adds an attention update only where its gate fires. The recurrent backbone and attention…
Block 0 receives the backbone's final normalized output. Blocks 1 and 2 apply their own pre-normalization. A shared, embedding-tied LM head supplies all three entropy evaluations and the final output prediction. A complete recurrent backbone, refined by three entropy-gated attention blocks. Looking for the larger model? AMOR-Gated DeltaNet 1.5B. AMOR (Adaptive Metacognitive Output Router) uses the model's predictive uncertainty to decide where attention should refine its recurrent representation. Each appended block reads normalized output entropy, compares it with a frozen threshold at inference, and adds an attention update only where its gate fires. The recurrent backbone and attention…
Block 0 receives the backbone's final normalized output. Blocks 1 and 2 apply their own pre-normalization. A shared, embedding-tied LM head supplies all three entropy evaluations and the final output prediction. A complete recurrent backbone, refined by three entropy-gated attention blocks. For a smaller download, see the 180M and 440M models below. AMOR (Adaptive Metacognitive Output Router) uses the model's predictive uncertainty to decide where attention should refine its recurrent representation. Each appended block reads normalized output entropy, compares it with a frozen threshold at inference, and adds an attention update only where its gate fires. The recurrent backbone and…
Block 0 receives the backbone's final normalized output. Blocks 1 and 2 apply their own pre-normalization. A shared, embedding-tied LM head supplies all three entropy evaluations and the final output prediction. A complete recurrent backbone, refined by three entropy-gated attention blocks. Looking for the larger model? AMOR-Mamba2 1.5B. AMOR (Adaptive Metacognitive Output Router) uses the model's predictive uncertainty to decide where attention should refine its recurrent representation. Each appended block reads normalized output entropy, compares it with a frozen threshold at inference, and adds an attention update only where its gate fires. The recurrent backbone and attention blocks…
Block 0 receives the backbone's final normalized output. Blocks 1 and 2 apply their own pre-normalization. A shared, embedding-tied LM head supplies all three entropy evaluations and the final output prediction. A complete recurrent backbone, refined by three entropy-gated attention blocks. Looking for the larger model? AMOR-Mamba2 1.5B. AMOR (Adaptive Metacognitive Output Router) uses the model's predictive uncertainty to decide where attention should refine its recurrent representation. Each appended block reads normalized output entropy, compares it with a frozen threshold at inference, and adds an attention update only where its gate fires. The recurrent backbone and attention blocks…
Anansi-35B-A3B balances good instruct following, narrative reasoning, and sweet prose. Specifically for my 8GB potato, because I can't run the ~30B dense models, but maybe it can be good for your potato too? This is the repo for Q8 prose/style head GGUFs. The Q8-Prose versions use the same Anansi LM output head — the DARE/TIES merge head interpolated with Melody — but restored at Q80 fidelity. The idea is to preserve more of the prose/style effect by avoiding some of the precision loss that comes from quantizing the head along with the rest of the model. There are also Q8-DarkScarlett versions using the same interpolated-head approach with ReadyArt's Dark Scarlett instead of Melody, giving…
Engineered by Deep Das • Part of the AntCoder Multi-Agent Coding Suite AntCoder-Builder-7B is a high-precision LoRA adapter fine-tuned on top of Qwen/Qwen2.5-Coder-7B-Instruct. It is specifically optimized to perform Contract-to-Implementation synthesis for complex, production-grade TypeScript applications. Given a strict TypeScript interface, class signature, function type contract, or JSDoc specification, AntCoder-Builder synthesizes the complete, strictly-typed implementation without type errors, missing properties, or hallucinated APIs. - Zero-Stub Completions (99.4%): Completely eliminates lazy // TODO, /... /, or throw new Error("not implemented") placeholders commonly emitted by…
Engineered by Deep Das • Part of the AntCoder Multi-Agent Coding Suite AntCoder-Fixer-7B is a specialized LoRA adapter fine-tuned on Qwen/Qwen2.5-Coder-7B-Instruct. It solves the single hardest problem in autonomous coding agents: hallucinatory patching and compile loops. Unlike generalist models that attempt to rewrite entire 500-line source files (frequently breaking unrelated functions or losing imports), AntCoder-Fixer acts as a surgical precision tool: 1. It ingests the exact TypeScript compiler diagnostic (e.g. TS2339: Property 'user' does not exist on type 'Session'). 2. It ingests the local 20-line source context window. 3. It emits a minimal, standard Git Unified Diff patch (…
The three-language base with a 2,048-token context: AnuLM-Base-400M continued for 10,000 steps at block 2,048 with YaRN, on 46M tokens of the same Hindi / English / Python proportions it was originally trained on. Five hours on one RTX 5070 Ti. Full log and the honest reading of what it bought: docs/RESULTS.md §28, with the zero-shot measurement it is compared against Hindi and English Wikipedia are CC BY-SA, C4 is ODC-BY, and the Python slice is codeparrot-clean, de-duplicated GitHub Python with mixed licences. Not affiliated with Sarvam AI, AI4Bharat, BharatGen or the Government of India. Every other checkpoint in this project is trained at 512 tokens. This one answers what happens if you…
A Qwen3.5 MoE decision model for choosing browser actions, selecting workflow steps, and judging natural-language criteria. Give the model a shared state and a set of candidate actions; the included decision runtime returns a distribution over those candidates. This repository contains 35B-A3B checkpoint-5949 merged BF16 weights. It is a standalone model with root-level Hugging Face configuration and weights, requiring no separate LoRA adapter. The native merged model scores 71/80 (88.75%) at its full 40-layer depth on the Frozen80 development panel. The decision interface accepts 2–16 request-specific candidates. Applications can use the returned candidate IDs to dispatch actions or build…
wikimedia/wikipedia - epfml/FineWeb2-HQ pipelinetag: text-generation - arabic - gpt2 - pretraining
This model is a fine-tuned version of fpadovani/arb-arab-100mb-ppt-Dp-100mbseed10. It has been trained using TRL. This model was trained with SFT.
This model is a fine-tuned version of fpadovani/arb-arab-100mb-ppt-shuff-dyck-100mbseed10. It has been trained using TRL. This model was trained with SFT.
This model is a fine-tuned version of fpadovani/arb-arab-100mb-ppt-shuff-dyck-10mbseed10. It has been trained using TRL. This model was trained with SFT.
The merged model is also here as GGUF, with the template and settings Ollama needs, so this one line runs it: Or build it from the Modelfile after downloading arety-phone-gemma4-12b-Q4KM.gguf and Modelfile: The GGUF has the adapter merged in, so no separate LoRA file is needed. It covers the text model only; Gemma 4's image and audio inputs are not included. 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
ONNX export of allenai/AstaBrief8B, published by Liodon AI. Exported with optimum (optimum.exporters.onnx.mainexport, task text-generation-with-past, so the graph exposes past-key-value inputs/outputs for KV-cached autoregressive decoding). Exported by Liodon AI
AstralLM is a compact, efficient causal language model built for Hugging Face Transformers. It features a clean decoder-only transformer architecture with Grouped Query Attention (GQA), SwiGLU feed-forward networks, RoPE positional embeddings, and per-head QK-norm — delivering strong performance at a small parameter count (~138M). AstralLM is a decoder-only transformer with the following design choices: The model uses 8 query heads and 4 key/value heads (2:1 ratio), halving the KV cache memory footprint during inference without measurable quality degradation. Each head operates over a head dimension of 80. Per-head RMSNorm is applied to both query and key projections before the rotary…
This model is a fine-tuned version of Qwen/Qwen2.5-1.5B-Instruct on the asynchowpythoncodeminutes dataset. It achieves the following results on the evaluation set: The following hyperparameters were used during training: - learningrate: 1e-05 - trainbatchsize: 1 - evalbatchsize: 1 - distributedtype: multi-GPU - gradientaccumulationsteps: 8 - totaltrainbatchsize: 8 - lrschedulertype: cosine - lrschedulerwarmupratio: 0.05 - numepochs: 2.0 - Transformers 4.57.6 - Pytorch 2.11.0+cu130 - Datasets 4.0.0 - Tokenizers 0.22.2
Due to system memory and time constraints, the figures shown below are from the conversion of GLM 5.1. We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.
This is a forge of pre-trained language models created using forgelm. This model was forged using the TIES forge method using Qwen/Qwen2.5-3B as a base. The following models were included in the forge: The following YAML configuration was used to produce this model
This is a forge of pre-trained language models created using forgelm. This model was forged using the SLERP forge method. The following models were included in the forge: The following YAML configuration was used to produce this model
LLM inference in C/C++ — https://github.com/ggml-org/llama.cpp
BAAR8-45M-Vision is an ultra-compact multilingual conversational model designed to enable real-time chat under minimal hardware conditions. With just 9 million (9M) parameters, it brings lightweight, interactive dialogue capabilities across 11 languages to extreme resource-constrained environments, embedded devices, and on-device platforms where standard LLMs cannot operate. BAAR8-45M-Vision은 최소한의 하드웨어 여건과 극단적인 리소스 제약 환경에서도 11개 언어로 실시간 대화(Chat)가 가능하도록 설계된 900만(9M) 파라미터 규모의 초경량 다국어 언어 모델입니다. 일반적인 대형 모델 구동이 불가능한 초저사양 엣지 및 임베디드 기기에서도 즉각적인 대화형 추론과 초경량 실시간 RAG 시스템 구축을 지원합니다. - Non-Commercial Use Only (비상업적 연구·개인 이용 한정): This model is strictly provided for non-commercial, research, and…
BaiHu-V1-Flash is an SSA (Sparse-attention + SubQ) retrofit of Qwen/Qwen3-0.6B-Base, fine-tuned on 1,294 bilingual multi-turn dialogues whose purpose is generalizing from a single worked example — the model is shown one worked rule / format / method in the first turn, and later turns ask it to reuse that rule on a new case it has never seen. Both revisions share the same SSA architecture, base weights and license. Every layer keeps the base model's MLP / RMSNorm weights and replaces full attention with an SSA layer built from three parallel paths: The previous revision compressed the whole prefix into a single mean vector. A query could only see one undifferentiated global average, the…
Conversational Writing Model｜口语化写作模型 Baihua-Gemma4-E4B 白话 Baihua 是Roland Wayne基于 Google Gemma 4 E4B-it 经过后训练的中文口语化写作模型。把已有内容，或整理好的事实、观点和提纲，写成像人类口述的正文，减少模板化、机械化的表达。 Roland Wayne 是内容创作者，也是白话 Baihua 的训练作者。在这个项目中，他负责确定表达目标、筛选训练材料，并通过实际测试和人工修改评估模型输出。 X：https://x.com/rwayne Github：https://github.com/ozrwayne - 微调方法：LoRA，rank 16，alpha 32；在前一轮风格 adapter 基础上继续进行配对微调。 - 完整权重：BF16 Safetensors，LoRA 已合并，加载完整权重时不需要再叠加同一 adapter。权重文件合计约 15.88 GB；含配套文件的完整包约 15.91 GB（十进制）。 - Ollama 性能测试使用约 14.94 GB 的 BF16 文本 GGUF 运行副本，归一化等少量张量为 FP32；它不包含纯文字推理不需要的音频和图像编码模块。原完整 Safetensors 模型保持独立。 - E4B 是基础模型的官方规格名称，不表示总参数只有 4B。官方标注约 4.5B 有效参数、含嵌入约 8B；当前 Ollama 文本副本实测为 7,463,013,674 个参数，约 7.5B。 - 配置总上下文：131,072…
Explains a Bash command line in exactly one English sentence, phrased as an instruction starting with a verb. QLoRA fine-tune of Qwen/Qwen2.5-Coder-3B-Instruct, merged and quantized to q4KM. The Modelfile pins the system prompt and this decoding: 920 held-out commands Judge columns are claude-sonnet-5 scoring a fixed 200-command subset. Apache-2.0, matching the base model, whose weights this GGUF contains. Training data comes from the nl2bash project; consult it for the terms attached to that corpus.
Explains a Bash command line in exactly one English sentence, phrased as an instruction starting with a verb. QLoRA fine-tune of Qwen/Qwen2.5-Coder-7B-Instruct, merged and quantized to q4KM. The Modelfile pins the system prompt and this decoding: 920 held-out commands Judge columns are claude-sonnet-5 scoring a fixed 200-command subset. Apache-2.0, matching the base model, whose weights this GGUF contains. Training data comes from the nl2bash project; consult it for the terms attached to that corpus.
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).
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).
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).
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).
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).
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).
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).
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).
libraryname: transformers pipelinetag: text-generation - language-model - berna - dna-kernel-plexus - cells - research - non-commercial 1.2B parameter language model with self-growing cell architecture. 2. Saturation S = 0.8578 (just above threshold 0.85) Comparison: GPT-2 Small (124M) achieves HellaSwag 31.4%, ARC-Easy 44.0%, PIQA 64.1%. Berna R7 v3 (1.2B) matches this at ~15x less training data. code/ Training code config/ config.json cellsruntimecurrent/ Live cells + cognitive log cellsruntimefinal/ Final cell state checkpoints/step00010200/ Final model (with wrapper) checkpoints/step00001700/ Earlier checkpoint data/tokenized/ eng.npy, math.npy, code.npy data/tokenizer/ tokenizer.json…
Each cell that saturates (S >= 0.68, K6 >= 0.03): 3. Gradient hooks - zero gradients for frozen neurons Result: Bounded Forgetting - old knowledge protected while learning new.
Bielik-11B-v3.0-Instruct post-trained for the Polish history matura (extended level) during the Warsaw Model Trainers alpha=64) on top of speakleash/Bielik-11B-v3.0-Instruct. closedem = exact match on closed items. Graded scores (deepseek-flash-high@think via benchmark/bench.py) are reported separately.
Bielik-11B-v3.0-Instruct post-trained for the Polish history matura (extended level) during the Warsaw Model Trainers alpha=64) on top of speakleash/Bielik-11B-v3.0-Instruct. closedem = exact match on closed items. Graded scores (deepseek-flash-high@think via benchmark/bench.py) are reported separately.
laptop / single GPU · Vision-capable Bonsai 2 27B — PrismML's ternary compression of Qwen3.8-27B — with the refusal circuitry surgically removed at the weight level while capability, vision, reasoning modes (off/low/xhigh), tool use, and multi-turn coherence are preserved. Full 27B-class hybrid Attention + SSM (GatedDeltaNet) architecture in a 5.9 GB 1bit GGUF. Proprietary weight-level abliteration by the dealignai research team. Byte-identical to the base ternary quant everywhere except a small set of tensors that carry the refusal circuit. Drop-in replacement for the ternary base at inference — same tokenizer, same chat template, same reasoning modes, same vision projector interface.…
repacked for the q27 engine. It is Qwen3.8-27B with every projection ternary (one fp16 scale per 128 weights) in a Hadamard-rotated basis, and no MTP block. This pack is an exact re-encoding of the PTQ10 GGUF, not a requantization: every trit is stored as-is in q27's 2-bit T2G128 container, the embedding and the head as int8 (also exact), and the rotation metadata is carried verbatim so the engine rotates activations itself. Needs q27 v0.12.0 or later (the MTP pack: v0.13.0) (docs/plans/2026-09-18-bonsai2-ternary.md in the repo has the design). The pack has no MTP head, so DFlash2 is the only speculative path. The Qwen3.8 drafter packs in work against this target as-is (3.47 tokens per…
Questions
Which Text generation models are most downloaded?
By monthly downloads reported by the Hugging Face Hub: .