Saved toolsSubmit a tool
FOXAI SEARCH

Search FoxAI

AI Models 2119

  • pplx-decider-v1.1-27b-GGUF↗

    Livesport

    llama.cpp port of perplexity-ai/pplx-decider-v1.1-27b, a decision model: it does not generate text, it answers typed questions about a shared state and returns calibrated probabilities. Same request shape as TypeSafe's Jev (`state` + `noul` / `choice` / `score` questions). The upstream checkpoint is bf16 safetensors that needs torch and ~49 GB of weights. This repo makes it runnable under llama.cpp. Both quants carry the decision head inside the GGUF as `cls.output.weight`; `readout.safetensors` is shipped as well so a runtime can load it independently.

  • DeepSeek-V4.1-Flash-1017-ds4-GGUF↗

    nazeshinjite

    A 2.36 bpw quant (excluding engrams) of DeepSeek-V4.1-Flash-1017 for DwarfStar (ds4), with 2.25bpw routed experts and 8-bit or higher everything else. It uses the same recipe and file layout as antirez's `DeepSeek-V4.1-Flash-Q2.gguf`, built from the 1017 weights with an imatrix calibrated on 1017. Designed for SSD-streaming on 128GB systems, where decode can reach up to 20tps with expert caching. Works in llama.cpp too, but I strongly recommend ds4. The language GGUF includes 188.83 GiB of native FP8 Engram tables. ds4 reads the rows it needs directly from disk and never loads the whole table, so keep the file on a fast local SSD. The language model loads in current ds4 with no changes. Memory options are the same as for antirez's Q2: SSD streaming on one 128 GB Mac, resident TP across two, or full residency on a larger Mac. See the ds4 model guide.

  • Qwen3.8-27B-EfficientThink-Uncensored-K3-Opus5-Grok4.6-GPT5.6Sol-SFT-SimPO-DFlash2-GGUF↗

    nerkyor

    **BF16 / FP8 main repository: nerkyor/Qwen3.8-27B-EfficientThink-Uncensored-K3-Opus5-Grok4.6-GPT5.6Sol-SFT-SimPO-DFlash2** Capability scores are shown only after a formal full-suite freeze, with every non-passing sample retained in the denominator. Q8_0, Q6_K, Q5-LynnStyle, Q4-LynnStyle, Q3-LynnStyle, and Q2-LynnStyle are published and have passed four-repository remote path, byte-size, and LFS-SHA256 readback. Lynn Agent v0.87.0 uses this release's **Q2-LynnStyle / Q3-LynnStyle + DFlash2** packages. The pairing passed runtime validation on DGX Spark; notarized Mac Apple Silicon and Intel builds, the Windows installer runtime check, CI in both repositories, matching `main` heads and tags across all three release repositories, complete SHA256 verification for 23 public files, and a remote CLI installation all passed.

  • pdmd_4NFE_lora↗

    pdmd2026

    This repository holds the **LoRA adapter** of a 4-step (4 NFE) student distilled from MiniMax-H3 with **Projected Distribution Matching Distillation (PDMD)**. The same model is also available as full transformer weights in pdmd2026/pdmd_4NFE_full. The adapter is applied to the transformer (`MiniMaxH3Transformer3DModel`) of the base model. Every other component (VAE, audio VAE, schedulers, text encoder, processor) is unchanged and comes from the base model. Tensor keys have the form `transformer. .lora_A.weight` / `transformer. .lora_B.weight`, where ` .weight` is the corresponding parameter of `MiniMaxH3Transformer3DModel`. Fusing rule (also stored in the safetensors metadata):

  • gemma-4-E2B-it-assistant↗

    Google

    > [!Note] > This model card is for the Multi-Token Prediction (MTP) drafters for the Gemma 4 models. MTP is implemented by extending the base model with a smaller, faster draft model. When used in a Speculative Decoding pipeline, the draft model predicts several tokens ahead, which the target model then verifies in parallel. This results in significant decoding speedups (up to 3x) while guaranteeing the exact same quality as standard generation, making these checkpoints perfect for low-latency and on-device applications. Gemma is a family of open models built by Google DeepMind. Gemma 4 models are multimodal, handling text and image input (with audio supported on E2B, E4B, and 12B) and generating text output. This release includes open-weights models in both pre-trained and instruction-tuned variants. Gemma 4 features a context window of up to 256K tokens and maintains multilingual support in over 140 languages. Featuring both Dense and Mixture-of-Experts (MoE) architectures, Gemma 4 is well-suited for tasks like text generation, coding, and reasoning. The models are available in five distinct sizes: **E2B**, **E4B**, **12B**, **26B A4B**, and **31B**. Their diverse sizes make them deployable in environments ranging from high-end phones to laptops and servers, democratizing access to state-of-the-art AI.

  • UlizaLlama↗

    Jacaranda

    UlizaLlama is a 7B Parameters language model that builds upon the foundation of Jacaranda/kiswallama-pretrained. Jacaranda/kiswallama-pretrained is a large language model continually-pretrained with 321,530,045 swahili tokens and a customized tokenizer with a swahili vocabulary of 20,000 tokens to extend the capabilities of Meta/Llama2. It offers significant improvements in both encoding and decoding for Swahili text, surpassing the Swahili performance of Meta/Llama2. Moreover, Jacaranda/kiswallama-pretrained excels in providing accurate next-word completions in Swahili, a capability which Meta/Llama2 falls short of. - Origin: Adaptation of the Jacaranda/kiswallama-pretrained model which is continually pretrained from Meta/Llama2. - Data: Instructional dataset in Swahili and English consisting of prompt-response pairs. - Training: Alignment to standard methodologies, incorporation of task-centric heads, neural network weight optimization via backpropagation, and task-specific adjustments. - Fine-tuning: Utilized the LoRA approach, refining two matrices that mirror the main matrix from Jacaranda/kiswallama-pretrained. This Low Rank Adapter (LoRa) was vital for instruction-focused fine-tuning. Post-training, the developed LoRa was extracted, and Hugging Face's merge and unload() function facilitated the amalgamation of adapter weights with the base model. This fusion enables standalone inference with the merged model Meanwhile, Jacaranda/kiswallama-pretrained) offers versatility in: - Text Summarization - Autoregressive Text Completion - Content Generation - Text Rewording - Grammar Refinement and Editing - Further Research-The current UlizaLlama is available as a 7 Billion parameters model, further research can also explore availing bigger variants of UlizaLlama. ## Sample Evaluation Prompts The use of the developed Large Language Model (LLM) capabilities is for research,social good and internal use purposes only. For commercial use and distribution, organisations/individuals are encouraged to contactJacaranda Health). To ensure the ethical and responsible use of UlizaLlama, we have outlined a set of guidelines. These guidelines categorize activities and practice

  • airoboros-70b-3.3↗

    jondurbin

    This model uses the llama-3-instruct prompt template, and is provided in the tokenizer config. You can use the `apply_chat_template` method to accurate format prompts, e.g.: BEGININPUT BEGINCONTEXT [key0: value0] [key1: value1] ... other metdata ... ENDCONTEXT [insert your text blocks here] ENDINPUT [add as many other blocks, in the exact same format] BEGININSTRUCTION [insert your instruction(s). The model was tuned with single questions, paragraph format, lists, etc.] ENDINSTRUCTION BEGININPUT BEGINCONTEXT date: 2021-01-01 url: https://web.site/123 ENDCONTEXT In a shocking turn of events, blueberries are now green, but will be sticking with the same name. ENDINPUT BEGININSTRUCTION What color are bluberries? Source? ENDINSTRUCTION

  • wav2vec2-lg-xlsr-en-speech-emotion-recognition↗

    ehcalabres

    The model is a fine-tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english for a Speech Emotion Recognition (SER) task. The dataset used to fine-tune the original pre-trained model is the RAVDESS dataset. This dataset provides 1440 samples of recordings from actors performing on 8 different emotions in English, which are: The following hyperparameters were used during training: - learning_rate: 0.0001 - train_batch_size: 4 - eval_batch_size: 4 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 8 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 - mixed_precision_training: Native AMP

  • Occamy-1.0-APEX-I-MiniPlus-V2.1-Abliterated-GGUF↗

    IsValorum

    > [!WARNING] > ### CRITICAL UPDATE (October 23, 2026): ABLITERATION RESTORATION & APOLOGY > **To the community:** I made a mistake and I apologize for it: I accidentally uploaded the base quantizations to the "abliterated" section, effectively duplicating the same GGUF files. This is why the abliteration appeared to work in my local tests, while the versions originally available on Hugging Face were actually just the base models. > > **Important Notice for Previous Downloads:** If you downloaded `Occamy-1.0.APEX-I-MiniPlus-V2.1-Abliterated.gguf` prior to October 23, 2026, you downloaded an un-abliterated (non-abliterated) build. > > **The resolution:** The repository has been fully corrected. The binary has been completely re-quantized using the authentic BF16 directional abliterated source with importance matrix (`imatrix`) calibration under the canonical 733-tensor APEX-I V2.1 recipe. Refusal suppression is verified active across all probe vectors. > > **Please re-download `Occamy-1.0.APEX-I-MiniPlus-V2.1-Abliterated.gguf` (14.75 GB / 13.74 GiB, verified with tensor-map.tsv).** > > Additionally, I have also added a companion **Occamy-1.0 APEX-I-NanoPlus-V2.1 Abliterated** version (12.54 GB / 11.68 GiB) to the official collection for memory-constrained workflows. If you consider these releases to be of high quality and they help you in your day-to-day workflow, consider supporting on Ko-fi. Every contribution directly sustains future cloud compute hours and helps support my university studies. > > *Independently computed on private cloud clusters. If this handcrafted release saves you VRAM and runs faster on your GPU, consider fueling the community compute fund on Ko-fi.* > > Explore the new canonical V3 releases. MiniPlus prioritizes Q5-class to near-Q6 fidelity without a fixed file-size ceiling; NanoPlus targets Q4-class quality at a practical footprint. > > - **Qwen3.8 27B — MiniPlus V3** — 14.16 GB **(+13.21% improvement in WikiText-2 PPL vs GSQ-RCO IQ3_S)**. > - **Qwen3.8 27B — NanoPlus V3** — 11.35 GB **(+11.49% improvement in WikiText-2 PPL vs GSQ-RCO IQ3_S)**. > - **Swift-1.5 Qwen3.8 27B Uncensored MTP — MiniPlus V3** — 14.28 GB **(+5.70% improvement in

  • Qwen3.8-27B-RTX5080↗

    ninfer-5080

    The model artifact itself remains byte-identical across the later runtime releases; v1.5 is the current production-qualified NInfer runtime/profile for this artifact. A matching SHA-256 identifies the exact validated project artifact regardless of filename or download machine. The same artifact has remained valid through later runtime optimization work. A newer NInfer runtime does **not** imply that a new `.ninfer` file is required.