サイト内検索

AI モデル 2119

• CodeNinja-1.0-OpenChat-7B↗

  beowolx

  CodeNinja is an enhanced version of the renowned model openchat/openchat-3.5-1210. It having been fine-tuned through Supervised Fine Tuning on two expansive datasets, encompassing over 400,000 coding instructions. Designed to be an indispensable tool for coders, CodeNinja aims to integrate seamlessly into your daily coding routine. CodeNinja maintains the same prompt structure as OpenChat 3.5. Effective utilization requires adherence to this format: The simplest way to engage with CodeNinja is via the quantized versions on LM Studio. Ensure you select the "OpenChat" preset, which incorporates the necessary prompt format. The preset is also available in this gist.

• Qwen3.8-Flash-Next-MLX-Serve-iQ-MLX-3.3bpw↗

  ddalcu

  mlx-serve pack of Qwen/Qwen3.8-Flash-Next, the Qwen4 preview architecture (`model_type: qwen4_exp`), quantized to 3.3 bits per weight with imatrix calibration so it fits a 64 GB Mac with room for 64k of context. 52 GB resident. Includes the MTP head and the vision tower (image and video input). Measured against the bf16 checkpoint on 1200 held-out next-token positions (agent, code, prose, math; documents the calibration never saw). Top-1 is how often the pack's greedy token equals bf16's; mass is the bf16 probability of the pack's pick, with bf16's own pick as the ceiling. Same positions, same server, cold prefills. The 3.3 bpw pack keeps 99% of the 4-8 pack's probability mass at 21 GB less.

• VieNeu-TTS-v3-Nano↗

  pnnbao-ump

  > [!WARNING] > **Preview release — still under active testing.** VieNeu-TTS v3 Nano is an experimental, lightweight model. It has known shortcomings (see *Limitations* below), its quality is **noticeably lower than v3 Turbo**, and weights, voices and the API may still change between revisions. **VieNeu-TTS v3 Turbo remains the default and recommended model.** Reach for Nano only when Turbo is too slow on your hardware or when you are deploying on a phone / edge device. **VieNeu-TTS v3 Nano** is a **48M-parameter flow-matching TTS model** for Vietnamese, built for **edge devices and weak CPUs**: old laptops, mini PCs, ARM single-board computers and Android phones. It ships as a handful of small **ONNX** graphs (≈280 MB at runtime), runs **torch-free on ONNX Runtime**, loads in about 3 seconds and synthesizes at roughly **3× the speed of v3 Turbo** on the same CPU. It is a different architecture from v3 Turbo (a non-autoregressive flow model instead of an LLM-style autoregressive backbone), which is where the speed comes from — and also where the quality gap comes from. It supports **6 built-in preset voices**, inline **emotion cues** (`[cười]`, `[thở dài]`, `[hắng giọng]`) and outputs **24 kHz** audio.

• revanimated↗

  stablediffusionapi

  Coding in PHP/Node/Java etc? Have a look at docs for more code examples: View docs payload = json.dumps({ "key": "", "model_id": "revanimated", "prompt": "actual 8K portrait photo of gareth person, portrait, happy colors, bright eyes, clear eyes, warm smile, smooth soft skin, big dreamy eyes, beautiful intricate colored hair, symmetrical, anime wide eyes, soft lighting, detailed face, by makoto shinkai, stanley artgerm lau, wlop, rossdraws, concept art, digital painting, looking into camera", "negative_prompt": "painting, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, deformed, ugly, blurry, bad anatomy, bad proportions, extra limbs, cloned face, skinny, glitchy, double torso, extra arms, extra hands, mangled fingers, missing lips, ugly face, distorted face, extra legs, anime", "width": "512", "height": "512", "samples": "1", "num_inference_steps": "30", "safety_checker": "no", "enhance_prompt": "yes", "seed": None, "guidance_scale": 7.5, "multi_lingual": "no", "panorama": "no", "self_attention": "no", "upscale": "no", "embeddings": "embeddings_model_id", "lora": "lora_model_id", "webhook": None, "track_id": None })

• TokForge-DreamShaper-LCM-GGUF-q4↗

  darkmaniac7

  A **Q4_0-quantized** single self-contained **GGUF** for `stable-diffusion.cpp`, packaging **DreamShaper-7** (SD1.5, realistic finetune) with the **LCM-LoRA** fused into the UNet for fast, guidance-free **few-step** sampling. This is the **6 GB-tier fast image route** for the TokForge apps, the smaller sibling of `TokForge-DreamShaper-LCM-GGUF` (f16, ~2.1 GB). It renders the same coherent people and hands (no SD-Turbo body-horror) at the same few-step LCM speed class, but at a smaller download + RAM footprint so it fits the 6 GB device tier where the f16 build is gated out. This GGUF was quantized with **stable-diffusion.cpp's own `-M convert --type q4_0`**, whose `tensor_should_be_converted` rule **protects embeddings and norms**; it keeps every tensor whose name contains `embedding` (and all `.bias` / `.scale` / norm tensors) at **F16**. So the CLIP text encoder's `token_embedding.weight` stays **F16**.

• TenEros_Max_optimized_learned↗

  LokkenJP

  Two accepted weight-only releases are available for the H3 Beta7 transformer from TenStrip/10Eros-Max. Both retain the source's embedded eight-step Turbo. **HQ (18.69 GB) remains the recommended choice for quality**, using learned W6 g16 and INT8 ConvRot weights. **Compact (14.65 GB)** adds W4 g16 to reduce the weight footprint for setups with less available memory. Both passed mechanical validation and operator ComfyUI acceptance on **2026-10-28**. For compact, the operator reports slightly lower quality than HQ while holding up well, with lower observed memory use. These are qualitative observations without measured VRAM savings or a controlled quality score. The older Beta6 compact remains a separate source release. Beta7 has the standard H3 tensor layout without compression-gate tensors. The upstream author describes its merge as MLP-focused, preserving base attention/reference behavior, and recommends Euler/simple or ER-SDE/SGM_uniform for its Turbo sampling. See the upstream source notes for its recipe, workflow settings and known limitations.

• sam-vit-tiny-random↗

  optimum-intel-internal-testing

• Qwen3.8-Flash-Next-REAP-288-MLX-4bit↗

  sh0wie

  Qwen3.8-Flash-Next with 288 of 512 experts per MoE layer, pruned with REAP saliency calibrated on the quantized weights, on the machine that serves them. It is a 180B-class MoE that runs on a 128 GB Mac for Claude Code-style agentic workloads and scores 91.5% on HumanEval. Decode is runtime-dependent: ~28 tok/s on stock mlx-vlm today, and 37 to 65 tok/s on the pmlx engine that ships with the tiered release (see "Decode speed" for the full breakdown). It supports MTP speculative decoding out of the box. Resident memory is ~68 GB as shipped, or ~39 GB when the n-gram table is streamed from NVMe (see "Running it"). For the ~39 GB streamed footprint, run under omlx v0.6.4 or newer. omlx offloads the n-gram table to NVMe automatically when the full model would not fit under the memory ceiling, so the build serves in about 39 GB resident. mlx-vlm runs the same build today at about 68 GB resident, with the n-gram table in memory. It streams to about 39 GB once PR #2045 (external PLE storage) lands in a release, since this repo already ships the `ple-store.json` manifest and `ple_storage` config that path reads. mlx-vlm needs `qwen4_exp` MTP support (git main after 2026-08-27, or any release that includes it):

• Qwen3-30B-A3B-GGUF↗

  6block

  All quants made using the imatrix option with a bilingual (English + Chinese) and code-heavy calibration dataset, so low-bit quants keep more of their Chinese and coding ability than a English-only calibration would. The `.imatrix` file used to produce these quants is included in this repo, so anyone can reproduce or extend the quant set with the exact same importance matrix. The first thing to figure out is how big a model you can run. For that you need to know how much RAM and/or VRAM you have.

• SNEX-V1↗

  gbuzhf

  **SNEX-V1 is Sage-A1.5 with a small, sparse transplant of Nex-N2.5-mini's training.** Sage is the body; from Nex only the largest changes its own training made are carried over, at a quarter of their strength. Everything else is Sage's: the vision projector, the MTP draft head, the router, the norms, the embeddings. Score: BF16 = 100; 70 % full-distribution fidelity `1/(1+KLD)` + 30 % same-top-1, averaged over the three sets. **1. Lineage first.** Weight comparisons placed the three models: Sage-A1.5 is a fine-tune of Qwen3.6-35B-A3B; Nex-N2.5-mini is a fine-tune of Qwen3.5-35B-A3B; Qwen3.6 continues Qwen3.5. So Nex's own training is `τ = Nex − Qwen3.5`. Taking the difference against Qwen3.6 instead would also subtract the 3.5 → 3.6 upgrade that Sage already carries.