サイト内検索
AI モデル 2119
- samantha-mistral-instruct-7b-GGUF↗
mradermacher
- Qwen2.5-7B-Instruct-1M-GGUF↗
Mungert
Our latest quantization method introduces **precision-adaptive quantization** for ultra-low-bit models (1-2 bit), with benchmark-proven improvements on **Llama-3-8B**. This approach uses layer-specific strategies to preserve accuracy while maintaining extreme memory efficiency. All tests conducted on **Llama-3-8B-Instruct** using: - Standard perplexity evaluation pipeline - 2048-token context window - Same prompt set across all quantizations **Key**: - PPL = Perplexity (lower is better) - Δ PPL = Percentage change from standard to DynamicGate - Speed = Inference time (CPU avx2, 2048 token context) - Size differences reflect mixed quantization overhead
- SlimSAM-uniform-77↗
Zigeng
> **0.1% Data Makes Segment Anything Slim** > Zigeng Chen, Gongfan Fang, Xinyin Ma, Xinchao Wang > Learning and Vision Lab, National University of Singapore > Paper: [[Arxiv]](https://arxiv.org/abs/2312.05284) > Code: [[GitHub]](https://github.com/czg1225/SlimSAM) **SlimSAM** is a novel SAM compression method, which efficiently reuses pre-trained SAMs without the necessity for extensive retraining. This is achieved by the efficient reuse of pre-trained SAMs through a unified pruning-distillation framework. To enhance knowledge inheritance from the original SAM, we employ an innovative alternate slimming strategy that partitions the compression process into a progressive procedure. Diverging from prior pruning techniques, we meticulously prune and distill decoupled model structures in an alternating fashion. Furthermore, a novel label-free pruning criterion is also proposed to align the pruning objective with the optimization target, thereby boosting the post-distillation after pruning. SlimSAM achieves approaching performance while reducing the parameter counts to **0.9\% (5.7M)**, MACs to **0.8\% (21G)**, and requiring mere **0.1\% (10k)** of the training data when compared to the original SAM-H. Extensive experiments demonstrate that our method realize significant superior performance while utilizing over **10 times** less training data when compared to other SAM compression methods.
- Swift-1.5-Qwen3.8-27B-Uncensored-NVFP4↗
ajgazin
NVFP4 checkpoint of ajgazin/Swift-1.5-Qwen3.8-27B-Uncensored-MTP, an abliterated Swift 1.5 Qwen3.8-27B (UkisAI's reasoning-efficient fine-tune of Qwen3.8-27B). For vLLM and SGLang. GGUFs for llama.cpp: Swift-1.5-Qwen3.8-27B-Uncensored-Dynamic-MTP-GGUF. Sampling, as for Swift and Qwen: temperature 1.0, top_p 0.95, top_k 20, min_p 0. The model thinks before answering by default. Same format, recipe, module split and tensor layout as the Swift 1.0 version, which was load-tested on an RTX 5090 with vLLM 0.29.0 (native W4A4, MTP decoding and image input working). This checkpoint was not load-tested separately.
- AuK-Base-and-Flash-GGUF↗
audio-cpp
These GGUF files are for use with audio.cpp. The card's `base_model` points to the Tencent AuK base model. This package also includes optional files converted from AuK-Flash; Flash is a separate distilled variant and is not the pinned base model. > **Output quality:** We validate parity with upstream Python for the tested configuration, not whether AuK's output meets every quality expectation. If a result is disappointing, listen to the corresponding Python reference WAV linked below first. A similar Python result points to the upstream model's behavior, not necessarily an audio.cpp conversion issue. The quantized component combinations have not received the same 16-task parity validation. Pass the downloaded directory to `audiocpp_cli --model`. Keep the `config/` and `tokenizer/` directories alongside the GGUF files. The default components are `auk-base-f32.gguf`, `qwen2.5-omni-3b-bf16.gguf`, and `auk-vae-f32.gguf`. AuK-Flash requires `--session-option auk.variant=flash`; its default generator is `auk-flash-f32.gguf`. The Qwen component is Qwen2.5-Omni-3B.
- JackOD-9B-Coder-MTP-GGUF↗
ManniX-ITA
GGUF quants of **`ManniX-ITA/JackOD-9B-Coder`**, a four-way `omnimerge_v2` merge over `Qwen/Qwen3.5-9B` built for **multi-turn agentic coding** — staying on a task across turns and tool calls until it is actually finished, and then stopping. > **v3 (2026-10-24).** v3 is the v2 recipe with the two smaller weights swapped: > **DeltaCoder 0.55 / Ornith-1.5-9B 0.30 / Coder-Flash 0.15** (v2: Coder-Flash 0.30 / Ornith 0.15). > `omnimerge_v2` density 0.53, darex-q 0.75, seed 42, the root `eos_token_id` 248046 / `pad_token_id` 248055 > fix and the Coder-Flash tokenizer and template are unchanged. v3 ships a **new serving sampler** and was > evaluated with it; the v2 and v1 results kept at the end of this card were measured **greedy**, so they are > a record, not a same-basis comparison. Every `_K` and `IQ` tier is **imatrix-calibrated**, including `Q6_K`. The calibration matrix used to build them is published in this repo as `imatrix.dat` (sha256 `feb286c4…`) so the quants can be audited and reproduced.
- InternVL2_5-2B↗
OpenGVLab
We are excited to introduce **InternVL 2.5**, an advanced multimodal large language model (MLLM) series that builds upon InternVL 2.0, maintaining its core model architecture while introducing significant enhancements in training and testing strategies as well as data quality. As shown in the following figure, InternVL 2.5 retains the same model architecture as its predecessors, InternVL 1.5 and 2.0, following the "ViT-MLP-LLM" paradigm. In this new version, we integrate a newly incrementally pre-trained InternViT with various pre-trained LLMs, including InternLM 2.5 and Qwen 2.5, using a randomly initialized MLP projector. As in the previous version, we applied a pixel unshuffle operation, reducing the number of visual tokens to one-quarter of the original. Besides, we adopted a similar dynamic resolution strategy as InternVL 1.5, dividing images into tiles of 448×448 pixels. The key difference, starting from InternVL 2.0, is that we additionally introduced support for multi-image and video data.
- pdmd_4NFE_full↗
pdmd2026
Full transformer weights (`MiniMaxH3Transformer3DModel`, bf16) of a 4-step (4 NFE) student distilled from MiniMax-H3 with **Projected Distribution Matching Distillation (PDMD)**. It is a drop-in replacement for the `transformer/` of the base model; every other component comes from the base model. The same model is also available as a LoRA in pdmd2026/pdmd_4NFE_lora. Sample with **4 denoising steps** using the base model's released scheduler configuration (shift 12 / 3).
- H3-Loras↗
poopooness
Mirrors of the LoRAs our MiniMax-H3 engine depends on, so production never depends on an upstream repo staying available. Verbatim copy of `larryvrh/MiniMax-H3-Turbo-Lora` (Apache-2.0), file `minimax_h3_turbo_v4_step600_ema.safetensors`. Not modified, not requantised, not re-keyed. Few-step audio-video sampling. 518 tensors, bf16, ~744 MB, rank 64 (16 on the AdaLN projections), applied as `W_eff = W + lora_B @ lora_A` with no alpha term.
- WaifuGemma4-26b-a4b-v1-i1-GGUF↗
hiwaifu-research
WaifuGemma4-26b-a4b-v1 is a role-play model: Gemma 4 26B-A4B (25.2B total, 3.8B active) post-trained with GRPO against a reward model learned from 1.2M blind votes cast by real users on the HiWaifu Arena. In blind head-to-head battles it splits votes evenly with GLM-5.1 and beats the untuned Gemma 4 26B-A4B-it 59.7% of the time. The full training story, arena results and response comparisons are on the model card. This repository holds the **weighted/imatrix** quantizations. Static quants and the vision projector (`mmproj`) are in WaifuGemma4-26b-a4b-v1-GGUF. Compared with the static quants, the imatrix versions cut the divergence from bf16 roughly in half at 4 bits: i1-Q4_K_M has a mean KLD of 0.069425 against 0.150663 for static Q4_K_M at the same 16.8 GB, and i1-IQ4_XS at 13.9 GB beats static Q4_K_M at 16.8 GB. Any recent llama.cpp works (the GGUFs were made with build `4fea119`, September 2026). The model was trained and evaluated in non-thinking mode; pass `--reasoning-budget 0` so the template inserts the empty thought channel the model expects.