Search FoxAI

AI Models 2119

• gemma-3-1b-it-unsloth-bnb-4bit↗

  unsloth

  Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3 models are multimodal, handling text and image input and generating text output, with open weights for both pre-trained variants and instruction-tuned variants. Gemma 3 has a large, 128K context window, multilingual support in over 140 languages, and is available in more sizes than previous versions. Gemma 3 models are well-suited for a variety of text generation and image understanding tasks, including question answering, summarization, and reasoning. Their relatively small size makes it possible to deploy them in environments with limited resources such as laptops, desktops or your own cloud infrastructure, democratizing access to state of the art AI models and helping foster innovation for everyone. These models were trained on a dataset of text data that includes a wide variety of sources. The 27B model was trained with 14 trillion tokens, the 12B model was trained with 12 trillion tokens, 4B model was trained with 4 trillion tokens and 1B with 2 trillion tokens. Here are the key components: The combination of these diverse data sources is crucial for training a powerful multimodal model that can handle a wide variety of different tasks and data formats.

• Yi-6B-200K↗

  01-ai

  If you're interested in Yi's adoption of Llama architecture and license usage policy, see Yi's relation with Llama. ⬇️ > 💡 TL;DR > > The Yi series models adopt the same model architecture as Llama but are **NOT** derivatives of Llama. 🔥 2024-07-29 : The Yi Cookbook 1.0 is released, featuring tutorials and examples in both Chinese and English.

• SAM3DBody-cpp-onnx-models↗

  AmmarkoV

• Latxa-Llama-3.1-70B-Instruct-v2-FP8↗

  HiTZ

  We introduce Latxa 3.1 70B Instruct, an instructed version of Latxa. This new Latxa is based on Llama-3.1 (Instruct), which we trained on our Basque corpus (Etxaniz et al., 2024) comprising 4.3M documents and 4.2B tokens using language adaptation techniques (paper in preparation). Our preliminary experimentation shows that Latxa 3.1 70B Instruct outperforms Llama-3.1-Instruct by a large margin on Basque standard benchmarks, and particularly, on chat conversations. In addition, we organized a public arena-based evaluation, on which Latxat competed against other baselines and proprietary models such as GPT-4o and Claude Sonnet. The results showed that Latxa ranked 3rd, just behind Claude and GPT-4 and above all the other same-size competitors. The official paper is coming soon. Latxa is a family of Large Language Models (LLM) based on Meta’s LLaMA models. Current LLMs exhibit incredible performance for high-resource languages such as English, but, in the case of Basque and other low-resource languages, their performance is close to a random guesser. These limitations widen the gap between high- and low-resource languages when it comes to digital development. We present Latxa to overcome these limitations and promote the development of LLM-based technology and research for the Basque language. Latxa models follow the same architecture as their original counterparts and were further trained in Latxa Corpus v1.1, a high-quality Basque corpora.

• Evoke-Turbo↗

  AlayaLab

  **Alaya-EVOKE-Turbo** is a **controllability-optimized** version of Alaya-EVOKE, focusing on camera-motion following and scene/subject instruction adherence. It uses the same architecture and inference workflow: **3 steps, no CFG, 384 × 640 at 24 fps**. Merged Turbo transformer weights, stored in **FP32** to match Evoke. Download the shared base components and depth backend separately: See Evoke for camera control, other inference modes, and long rollouts.

• Qwen3.6-27B-Uncensored-HauhauCS-Aggressive↗

  HauhauCS

  > **Join the Discord** for updates, roadmaps, projects, or just to chat. > **Not sure which variant to pick?** 99.9%+ of users should use **Balanced** — same 0/465 refusal rate, more stable sampling, great for agentic coding / tool-use / reasoning / creative writing. Pick **Aggressive** only if you specifically want the model to skip its preamble on hardcore prompts. > **HuggingFace's "Hardware Compatibility" widget doesn't recognize K_P quants** — it may show fewer files than actually exist. Click **"View +X variants"** or go to **Files and versions** to see all available downloads.

• Qwen3.8-27B-W4A16-AWQ↗

  philbert440

  INT4 weight-only quantization of Qwen/Qwen3.8-27B — **the whole model, nothing stripped**: vision tower intact, MTP speculative-decoding head intact, thinking mode calibrated with real reasoning traces. 55 GB BF16 → **19.5 GB**, which puts a current-generation 27B VLM (with headroom for 32K+ context) on a pair of 32 GB V100s — and it serves on modern GPUs with stock vLLM, no ceremony. Validated end-to-end on **2×V100-32GB under 1Cat-vLLM 1.2.2** (SM70 TurboMind dequant path): coherence, think-tag discipline, image understanding, math, long-form generation, and MTP speculative decoding in both draft modes. **Methodology:** warm serve (2 discarded warmup generations), fixed-length generations via `ignore_eos` so every run produces exactly the stated token count, official Qwen3.8 sampling per mode (thinking `1.0/0.95/20`, instruct `0.7/0.80/20 + presence 1.5`), varied prompts. Reported as mean ± sd tokens/s. Rig: 2×V100-32GB, 1Cat-vLLM 1.2.2, TP2, `fp8_e5m2` KV, `max_num_seqs 4`, MTP K=2.

• msmarco-MiniLM-L6-en-de-v1↗

  cross-encoder

  This is a cross-lingual Cross-Encoder model for EN-DE that can be used for passage re-ranking. It was trained on the MS Marco Passage Ranking task. The performance was evaluated on three datasets: - **TREC-DL19 EN-EN**: The original TREC 2019 Deep Learning Track: Given an English query and 1000 documents (retrieved by BM25 lexical search), rank documents with according to their relevance. We compute NDCG@10. BM25 achieves a score of 45.46, a perfect re-ranker can achieve a score of 95.47. - **TREC-DL19 DE-EN**: The English queries of TREC-DL19 have been translated by a German native speaker to German. We rank the German queries versus the English passages from the original TREC-DL19 setup. We compute NDCG@10. - **GermanDPR DE-DE**: The GermanDPR dataset provides German queries and German passages from Wikipedia. We indexed the 2.8 Million paragraphs from German Wikipedia and retrieved for each query the top 100 most relevant passages using BM25 lexical search with Elasticsearch. We compute MRR@10. BM25 achieves a score of 35.85, a perfect re-ranker can achieve a score of 76.27. We also check the performance of bi-encoders using the same evaluation: The retrieved documents from BM25 lexical search are re-ranked using query & passage embeddings with cosine-similarity. Bi-Encoders can also be used for end-to-end semantic search.

• regnet-y-040↗

  facebook

  RegNet model trained on imagenet-1k. It was introduced in the paper Designing Network Design Spaces and first released in this repository. Disclaimer: The team releasing RegNet did not write a model card for this model so this model card has been written by the Hugging Face team. The authors design search spaces to perform Neural Architecture Search (NAS). They first start from a high dimensional search space and iteratively reduce the search space by empirically applying constraints based on the best-performing models sampled by the current search space.

• Llama-3.2-1B-Instruct-TOFU-forget10-GradDiff-SAM↗

  Lordvarun23