Open-Weight Models Under Apache License 2.0

Model · Feature extraction

univ_conformer_small_dr_vectors

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Feature extraction

univ_conformer_small_ext_softmax

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Image and text to text

GameScaling-2B-S3

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-2B-S3. Qwen3.5-2B continued pre-trained on a balanced mixture of 27% Minecraft and other games (S3), 1,600 h of gameplay, 6,760 training steps; 5.4 GB of bfloat16 weights on 1 GPU. Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-2B-S3 8000 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms each, plus one mouse…

Model · Image and text to text

GameScaling-9B-S3

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-9B-S3. Qwen3.5-9B continued pre-trained on a balanced mixture of 27% Minecraft and other games (S3), 1,600 h of gameplay, 6,760 training steps; 18.8 GB of bfloat16 weights on 1 GPU. Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-9B-S3 8000 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms each, plus one mouse…

Model · Feature extraction

ls_conformer_small_sphereface

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Feature extraction

ls_transformer_small_ge2e

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Text generation

Qwen3.8-27B-pi-mlx-6Bit

SolsticeAI

The Model SolsticeAI/Qwen3.8-27B-pi-mlx-6Bit was converted to MLX format from bytkim/Qwen3.8-27B-pi using mlx-lm version 0.31.2.

A distilled fine-tune of the Qwen3.6-35B-A3B base (MoE, 3B active), trained on data generated by Qwen3.8-27B as the teacher — bringing the base model to (and beyond) the teacher's capability while keeping the MoE speed advantage. - LoRA adapter (40 MB, mount on the base, saves 22 GB) — lododoai/Qwen3.6-35B-A3B-Distill-Qwen3.8-27B-GGUF-lora - Community work, not affiliated with or endorsed by Qwen. Base model: Qwen3.6-35B-A3B (Apache-2.0). moefast is a small launcher that configures llama.cpp for your machine. It reads the model's GGUF metadata, measures your free VRAM, computes the expert-cache size, context and batch parameters, writes them into an editable config file next to the model…

Binary classification model that predicts whether EUR/USD will close higher (UP) or lower (DOWN) the next trading day. The model uses 53 features including: - Williams %R, CCI: Additional momentum indicators See predict.py for a complete inference example. Based on published financial ML literature: This model is for research and educational purposes only. It is NOT financial advice. Forex trading involves significant risk. Past performance does not guarantee future results. Realistic accuracy for daily direction prediction is 52-56% (literature consensus). This model repository was generated by ML Intern, an agent for machine learning research and development on the Hugging Face Hub.

Model · Text generation

kas-4b

Kashav Piya

A fine-tuned decision engine based on Qwen/Qwen3-4B, submitted to the Decision Index v0.2.1 leaderboard. Decision Index: 40.1 (raw 54.18) — above Tev1-4B (29.2), within 6 points of Scion v4 9B (45.9). Per-area breakdown against all 4B submissions on Decision Index 0.2.1. kas-4b leads the 4B tier on: - Knowledge & Reasoning (39.5) — highest of all 4B models; higher LoRA rank (r64) likely helps on harder reasoning tasks - Arts & Human Taste (37.7) — highest of all 4B models by a wide margin - Language Understanding (35.2) — 13–25 points behind ezjev and Nox; language-heavy fine-tuning data favors those models - Tools & Automation (49.0) — behind ezjev (69.9) and Nox (60.1) Profile: the most…

Model · Feature extraction

ls_mamba_small_ge2e

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Token classification

physics-ai999x

AI999X Labs

License: Apache-2.0 · Author: Ibnul Jaif Farabi (creator and engineer) · Organization: AI999X · Version 2.0.0 physics-ai999x is an offline, deterministic Python solver that answers physics questions with verified results from 148 certified capabilities, gives 859 documented answers identically on Windows and Linux, and refuses rather than guesses when an answer cannot be verified. physics-ai999x is a compact, deterministic physics engine for Python. It reads a question in plain English (or a typed call), picks a certified physics capability, computes the answer exactly, checks it, and returns the result with its steps, units, LaTeX and assumptions — or an honest refusal that says what is…

Model · Image and text to text

GameScaling-0.8B-S1

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-0.8B-S1. Qwen3.5-0.8B continued pre-trained on Minecraft gameplay only (S1), 1,600 h of gameplay, 6,736 training steps; 2.2 GB of bfloat16 weights on 1 GPU. Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-0.8B-S1 8000 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms each, plus one mouse movement and scroll for the…

Model · Feature extraction

ls_conformer_small_pfas

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Image and text to text

GameScaling-2B-S1

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-2B-S1. Qwen3.5-2B continued pre-trained on Minecraft gameplay only (S1), 1,600 h of gameplay, 6,736 training steps; 5.4 GB of bfloat16 weights on 1 GPU. Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-2B-S1 8000 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms each, plus one mouse movement and scroll for the whole…

Model · Feature extraction

univ_conformer_base_ge2e

Quan

This model was trained for 400 gradient steps with mini-batches of 32 utterances (8 speakers x 4 utterances), i.e. about 1.35 epochs over the 9,447-utterance universal corpus, on a CPU workstation, with a single seed (42). Losses are not converged at this budget. All EERs below are therefore far from state of the art and are only meaningful for relative comparisons between FlaxSpeaker checkpoints trained under the same budget. Training-crop provenance. This checkpoint was trained with fixed 180-frame (1.8 s) crops and all-ones validity masks, not with the U(90, 180) variable-length crops that are the current pipeline default: the training code in use at the time defaulted the batchers to…

Model · Image and text to text

GameScaling-27B-S3

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-27B-S3. Qwen3.5-27B continued pre-trained on a balanced mixture of 27% Minecraft and other games (S3), 1,600 h of gameplay, 7,020 training steps; 54.7 GB of bfloat16 weights on 2 GPUs (tensor parallel). Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-27B-S3 8000 2 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms…

Model · Token classification

algebra-ai999x

AI999X Labs

License: Apache-2.0 · Author: Ibnul Jaif Farabi (creator and engineer) · Organization: AI999X · Version 2.0.0 algebra-ai999x is an offline, deterministic Python solver that answers algebra questions with verified results from 56 certified capabilities, gives 230 documented answers identically on Windows and Linux, and refuses rather than guesses when an answer cannot be verified. algebra-ai999x is a compact, deterministic algebra engine for Python. It reads a question in plain English (or a typed call), picks a certified algebra capability, computes the answer exactly, checks it, and returns the result with its steps, units, LaTeX and assumptions — or an honest refusal that says what is…

Model

szl-kernels

SZL Holdings

Control before action. Evidence after. Part of the szl-holdings estate · The protected main workflow verifies the kernel suite, replays the MiniEmbed artifact within its declared tolerance, and checks the source-binding contract against the public Hugging Face artifact. Exact publication is performed by the authorized release gateway in That gateway checks out an exact protected Git revision, publishes the declared file set, and reads every byte back at the resulting immutable Hub revision. The exact Git source revision is written to publication.json. This is a governed kernel suite with a receipted word-embedding companion. It is not a general-purpose language model, and its intrinsic…

Model · Image and text to text

GameScaling-9B-S1

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-9B-S1. Qwen3.5-9B continued pre-trained on Minecraft gameplay only (S1), 1,600 h of gameplay, 6,736 training steps; 18.8 GB of bfloat16 weights on 1 GPU. Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-9B-S1 8000 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms each, plus one mouse movement and scroll for the…

weighted/imatrix quants of https://huggingface.co/darkc0de/RICO-v2 For a convenient overview and download list, visit our model page for this model. static quants are available at https://huggingface.co/mradermacher/RICO-v2-GGUF This is a vision model - mmproj files (if any) will be in the static repository. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant And here are Artefact2's thoughts on the matter…

Model · Text classification

Laya

AXERA

Ready-to-run deployment package for the Laya typed-decision model family on AX650 / NPU3. Laya is a bidirectional decision model. It evaluates user-defined questions over text or structured JSON state without generating text. - [x] NPU3 The packaged runtime does not automatically route between checkpoints. Select the directory that matches the input language or workflow. Users submit a JSON object with two fields: - state: the text or structured record to evaluate, such as a support ticket, security incident, email, or agent trace. - questions: named decision definitions. Each question contains a decision type, an instruction, and the candidate criteria when applicable. The runtime supports…

Model · Text generation

StandardOne-3B-FP8

Standard Thinking

StandardOne-3B-FP8 is an FP8 (compressed-tensors, float8e4m3 weights, dynamic per-token activations) quantization of the released StandardOne-3B decision model. The language-model linear projections (q/k/v/o, gate/up/down) are quantized per-channel FP8 E4M3 with dynamic FP8 activations (llm-compressor's data-free FP8DYNAMIC recipe, no calibration data required); the vision tower, multi-modal projector, embeddings and lmhead are left unquantized in BF16. It was produced from source revision 68dafd17ead9b8cf6f85c4f08f7f2f3a1e7b9e5c of StandardOne-3B on 2026-09-25 using llm-compressor 0.14.0 (torch 2.14.0, transformers 5.17.0, compressed-tensors 0.19.0); results below. Served through SGLang…

Model · Image and text to text

GameScaling-27B-S1

Tsinghua Spatial Intelligence Lab

This repository: GameScaling-27B-S1. Qwen3.5-27B continued pre-trained on Minecraft gameplay only (S1), 1,600 h of gameplay, 6,995 training steps; 54.7 GB of bfloat16 weights on 2 GPUs (tensor parallel). Serve it with bash scripts/servevllm.sh THUSI-Lab/GameScaling-27B-S1 8000 2 (§1). The inference client, prompts and scripts are in THUSI-Lab/GameScaling; all eight checkpoints are in the GameScaling models are Qwen3.5 vision-language models with continued pre-training on keyboard-and-mouse gameplay. Each call takes the current game frame (plus a short history of past frames and actions) and returns the next 200 ms of keyboard and mouse actions: 6 steps of 33 ms each, plus one mouse movement…

Part of the Qwen3-VL-8B Hebrew / Cairo Genizah fine-tune series (https://github.com/AIStream-Peelout/historical-document-analysis). Flagship: v21b step 1200. Status: former flagship (August 2026), superseded by v20a and then v21b. Kept as the reference point the later cards compare against. Trained 2026-08-20/21, 1,300 steps (the run stopped at ~1,310); warm start from v18b step 700 (weights only). pages, prompted regions, sections and line crops) — and synthetichebrewv3 (12k renders over twelve faces: Solitreo, Rashi, STA"M and square-letter drills). Carried over: genizahcleanv2 (1,055 curated documentary pages with published transcriptions) and a small Talmud replay. Mixture: KTIV…

Official repository for gap-efficient-sam checkpoints, evaluated on AI-generated image detection and segmentation tasks. To download the latest checkpoint directly to your local workspace

Model · Text generation

StandardOne-8B-FP8

Standard Thinking

StandardOne-8B-FP8 is an FP8 (compressed-tensors, float8e4m3 weights, dynamic per-token activations) quantization of the released StandardOne-8B decision model. The language-model linear projections (q/k/v/o, gate/up/down) are quantized per-channel FP8 E4M3 with dynamic FP8 activations (llm-compressor's data-free FP8DYNAMIC recipe, no calibration data required); the vision tower, multi-modal projector, embeddings and lmhead are left unquantized in BF16. It was produced from source revision e88423700bb5ab9b2f50e176cf19825914345272 of StandardOne-8B on 2026-09-25 using llm-compressor 0.14.0 (torch 2.14.0, transformers 5.17.0, compressed-tensors 0.19.0); results below. Served through SGLang…

Model · Token classification

astrlink-guard

QuantumNous

AstrLink Guard is a lightweight model for detecting personal information and credentials in Chinese, English and mixed Chinese–English text, including source code, configuration files, logs and coding-assistant requests. It runs fully locally on CPU with ONNX Runtime and returns entity spans with labels and scores, so the calling application decides what to mask. It is the privacy detection model used by AstrLink. This repository contains version 0.1.1 in two formats: int8/ (dynamic INT8, recommended) and fp32/ (FP32, higher accuracy). - Fewer credentials left unmasked. At the recommended minimum score of 0.8, credentials in coding-assistant requests that are missed with no part masked drop…

A full fine-tune of OpenGVLab/InternVL35-2B-HF (vision tower, projector and LLM all trained) on rejection-sampled, verified answers from public biomedical datasets. No MMBU data was used, and every training image passes a pHash near-duplicate check against the MMBU public images. - RFT data. The base model answered public classification, fine-grained, detection and counting questions 6 times each. Only answers that verified against the ground truth were kept. Hint-rationalized answers were added for unsolved questions, and box-answer rows were excluded. - Metadata supervision (about 30% of the mix). Targets are modality, submodality, specimen, body part, stain and medical domain. Medical…

Model · Zero-shot classification

matilda-jev-v1.5

Maincode

A BF16 JEV decision model with a 255-option readout. This is the highest-scoring checkpoint in the completed MATILDA JEV experiments as of 9 October 2026: Decision Index 63.47 (0.3 public suite). The model accepts a state and one or more decision questions. It supports choice, noul (probability of true), and ordinal score. Inference is one forward pass per question; the model has no text-generation head and does not generate a chain of thought. The LoRA updates are already merged into the weights. No adapter merge is required. Use Python 3.12 or newer and install PyTorch for your accelerator. The validated cluster environment uses PyTorch 2.14.0, Transformers 5.17.0, and an AMD Instinct…

Model · Image and text to text

Qwen3.5-9B-onw

Ryu Gyo

Qwen/Qwen3.5-9B を onw（俺のNPUがこんなに動くわけない）用に変換したモデルです。onw は LLM を Intel NPU だけで動かすエンジンで、このリポジトリにはモデルだけが入っています（エンジンは別にダウンロードします）。 テキストと画像に対応。Hugging Face の重みから onw convert で直接作っています（単独版は ryugyosoft/Qwen3.5-9B-npu）。 私たちの知る限り、Gated DeltaNet 混在のハイブリッドモデルを画像入力まで含めて NPU で動かす公開実装は onw が初めてです（2026 年 9 月、Fable 5.1 調べ。OpenVINO 公式は Qwen3.5 を CPU・GPU のみ対応としています）。 1. onw（エンジン）を入れます。 まだの場合は、onw の「インストール」 のとおり 1 行貼るだけです。Windows はスタートメニューで「PowerShell」を開き、次の 1 行を貼って Enter（Ubuntu は onw の README の curl... | bash）。 2. タスクトレイの onw アイコンから開く onw ウィンドウ の「モデル」タブで Qwen3.5-9B を選び、「ダウンロード」を押します（5.3 GB）。 3. 「サーバー」タブでモデルを選んで「ロード」を押します。初回だけ NPU 向けの準備に約 13 分かかります。 4. 「チャット」ですぐ試せます。他のアプリからは OpenAI 互換 API として使えます（base URL は…

Model · Tabular regression

canrun-fps-predictor

Bartosz Lenart

A small tabular model that estimates the native frame rate of a PC game at 1080p, 1440p or 4K from two things you can know before buying or installing: the PC's hardware scores and the game's Steam store requirements. It is the FPS engine of the CanRun G-Assist plugin and runs with the Python standard library alone. Real-world accuracy: on independent benchmarks the model alone is a median 18.9% off on recent games and NVIDIA cards, and 37% to 39% off on other test sets. The full CanRun app, which adds benchmark curves and live Steam store lookups, is 15.2% off end to end. No Steam player FPS data is used anywhere. Treat every prediction as a rough estimate, good for "will this run at 60…

Model · Image and text to text

Qwen3.5-397B-A17B-CQ-4bit

ThousandML

Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance. Qwen3.5 represents a significant leap forward, integrating breakthroughs in multimodal learning, architectural efficiency, reinforcement learning scale, and global accessibility to empower developers and enterprises with unprecedented capability and efficiency. For more details, please refer to our blog post Qwen3.5. WMT24++: a harder subset of WMT24 after difficulty labeling and rebalancing; we report the averaged scores on 55 languages using XCOMET-XXL. Empty cells (--) indicate scores not yet available or not applicable. MathVision：our model’s score is…

Model · Image and text to text

Qwen3.5-4B-W4A16-AutoRound-NVFP4

Vishva R

This is an NVFP4 (W4A16: 4-bit weight, 16-bit activation) quantized version of Qwen/Qwen3.5-4B, produced using AutoRound — Intel's sign gradient descent based quantization method designed for production-grade accuracy retention. Exported via compressed-tensors (llmcompressor format) with MTP layers preserved. - NVFP4 Quantization — Weights are quantized to FP4 micro-scaled formats using group size 16; activations remain in 16-bit for stable numerical fidelity. - Vision Tower (visual / quantnontextmodule): Kept in BF16 to preserve visual reasoning and OCR precision. - Special Modules (ignorelayers & layerconfig): linearattn, mtp, and mtp.fc are strictly retained in native bfloat16 to ensure…

Model · Image and text to text

Qwen3.5-9B-W4A16-AutoRound-NVFP4

Vishva R

This is an NVFP4 (W4A16: 4-bit weight, 16-bit activation) quantized version of Qwen/Qwen3.5-9B, produced using AutoRound — Intel's sign gradient descent based quantization method designed for production-grade accuracy retention. Exported via compressed-tensors (llmcompressor format) with MTP layers preserved. - NVFP4 Quantization — Weights are quantized to FP4 micro-scaled formats using group size 16; activations remain in 16-bit for stable numerical fidelity. - Vision Tower (visual / quantnontextmodule): Kept in BF16 to preserve visual reasoning and OCR precision. - Special Modules (ignorelayers & layerconfig): linearattn, mtp, and mtp.fc are strictly retained in native bfloat16 to ensure…

Model · Speech recognition

qwen3-asr-1.7b-bf16-mlx

App Automaton

MLX-native bf16 conversion of Qwen3-ASR-1.7B for local automatic speech recognition on Apple Silicon — English, Chinese, and mixed Chinese/English. It runs through the mlx-speech runtime with no PyTorch and no cloud API at inference time. Weights ship as plain.safetensors. Install mlx-speech, then load by alias (downloads on first use): Or download the weights once and load by local path: - This repo contains the MLX runtime artifact only (no PyTorch checkpoint). - v0 is single-pass offline ASR; streaming, timestamps, and long-audio chunking are deferred. - For a smaller download, use the int8 build. Apache 2.0, following the upstream Qwen3-ASR release. Refer to the original Qwen release…

Model · Speech recognition

sea-asr-id

Pixel ML

Pixel ML builds video intelligence for Southeast Asian media. These are the speech models we run for Southeast Asian media: captions, search, clips. Open weights, open data, runs on your GPU — or on a CPU. Google's EmbeddingGemma 2 audio encoder with a CTC head, trained by Pixel ML on openly licensed Indonesian speech (YODAS, FLEURS, Common Voice). An optional n-gram language model (folder lm/, 0.27 GB) built from Indonesian Wikipedia comes with the weights. Speed (one H100 80GB, side by side with Whisper on the same GPU type and container size; seaasr v0.2.1): greedy decoding is about 18x faster than batched Whisper large-v3 and 9x faster than large-v3-turbo on a 79-minute recording…

Model · Zero-shot classification

finecat-nli-v2-xl

Lee Miller

This model is a fine-tune of jhu-clsp/ettin-encoder-1b, trained on the dleemiller/FineCat-NLI-Reason dataset, a compilation of several high-quality NLI data sources with quality screening and reduction of easy samples in the training split. Also includes additional NLI, fact verification and long document sources (2.05M pairs in total). The training uses logit distillation from an NLI finetune of Qwen3.8-27B. \begin{equation} \mathcal{L} = \alpha \cdot \mathcal{L}{\text{CE}}(z^{(s)}, y) + \beta \cdot T^2 \cdot \mathrm{KL}\left(\sigma(z^{(t)}/T) \,\|\, \sigma(z^{(s)}/T)\right) \end{equation} where \\(z^{(s)}\\) and \\(z^{(t)}\\) are the student and teacher logits, \\(y\\) are the ground…

Model · Text generation

ProactiveInquirer-Qwen3-8B

Ido Levy

Ido Levy 1,2 · 1,2 This is the trained questioner from Asking for What Was Never Requested: Horizontal and Vertical Proactivity in Agents. It is a LoRA adapter on Qwen3-8B, trained with Q&D (questioner and drafter). A tool-using agent usually does what it is asked, yet a task often needs information the user never mentions. This model decides, one step at a time, which question to send to the agent's retriever next, or that it is time to stop. It goes after two kinds of unstated need: ZIP code, so the agent looks up the account. the order names the product, and the product lists the size-8 variant. These numbers are from the paper, on held-out test splits. The MuSiQue reading compares…

Version v2.0 (2026-10-08). main is this build. Version v1.0 (T-cloq16, 2026-10-06) stays on the v1.0 tag. Unofficial post-training quantization of Cloudflare/clef-flash (revision 17f0b0ad64efb65d273590632833508766b2aae6). This is not an official Cloudflare release and is not endorsed by Cloudflare or the Qwen team. Licensed Apache-2.0 like the original (see LICENSE, NOTICE.md for the changes). Text-only package: the vision tower is removed. The separate vision repo is unchanged. Paired cluster bootstrap of accuracy, v2.0 − v1.0, 10,000 resamples, seed 1234, 95% percentile CI, clusters = (source, group) within source. Same items, same order, existing per-question rows (no re-run). - Backbone…

Model · Text classification

laya-multilingual-calculator-router

Leo Borcherding

A Laya decision model that answers one question before each agent step: does this step need a calculation? It also scores how much working the calculation takes (none, trivial, real working). Fine-tuned from convaiinnovations/laya (the multilingual checkpoint) on Borcherding/calculator-routing with Unsloth's FastDecisionModel. Ask the questions with the exact wording above; that is what the model was trained on. Held-out test split (1,200 rows, 600 per class). Test uses templates, names and entities that never appear in train, so it measures new phrasings of familiar kinds of step. Base Laya says "yes" to almost anything with a number in it. The tuned model fixes that: order IDs, versions…

INT4 weight-only GPTQ quantization of Qwen/Qwen3.5-4B for memory-constrained single-GPU serving. Also the verifier for the (DFlash2 speculative decoding, τ 3.8 at the serving point / 4.1 greedy vs native MTP 2.9 / 3.0 on the same ladder). - W4A16 (INT4 weights, BF16 activations), group 128, act-order, compressed-tensors (pack-quantized) — vLLM Marlin kernels - Calibrated with llm-compressor dataset-native proportions from (~2.6M effective tokens). The exact rows are hash-pinned in — that pin is the citable calibration generation. (A -cascade suffix on local paths in the lab repo refers to the same generation; this repo carries only the current one and supersedes the initial pre-cascade…

Model · Image and text to text

gemma-4-E4B-it-onw

Ryu Gyo

google/gemma-4-E4B-it を onw（俺のNPUがこんなに動くわけない）用に変換したモデルです。onw は LLM を Intel NPU だけで動かすエンジンで、このリポジトリにはモデルだけが入っています（エンジンは別にダウンロードします）。 テキストと画像に対応。単独版 ryugyosoft/gemma-4-E4B-it-npu と同じ検証済みの NPU グラフを、共通エンジンで動かします。 私たちの知る限り、画像エンコーダーと層ごとの埋め込みまで含めて Gemma 4 全体を NPU で動かす公開実装は onw が初めてです（2026 年 9 月、Fable 5.1 調べ。Intel 公式の OpenVINO GenAI は画像エンコーダーを CPU で動かし、他の公開例はテキストのみです）。 1. onw（エンジン）を入れます。 まだの場合は、onw の「インストール」 のとおり 1 行貼るだけです。Windows はスタートメニューで「PowerShell」を開き、次の 1 行を貼って Enter（Ubuntu は onw の README の curl... | bash）。 2. タスクトレイの onw アイコンから開く onw ウィンドウ の「モデル」タブで gemma-4-E4B-it を選び、「ダウンロード」を押します（4.3 GB）。 3. 「サーバー」タブでモデルを選んで「ロード」を押します。初回だけ NPU 向けの準備に約 4 分かかります。 4. 「チャット」ですぐ試せます。他のアプリからは OpenAI 互換 API…

Unless you are studying this series' training history, use the flagship instead: qwen3-vl-8b-hebrew-v19a-ckpt. Status: superseded (historical). The Rashi-script generation: warm-started from the v1 winner and trained on a mixture led by synthetic Rashi-script data (40%) plus Talmud small-script and gemara material, at a 3.2 MP resolution floor. on a held-out synthetic Rashi eval slice whose text is unmemorizable, i.e. the model reads glyphs rather than reciting canonical text. Also solved crop-level Vilna gemara (CER 0.019). - Talmud page-level (strict CER means, 65-page benchmark): gemara 0.059 / rashi 0.362 / tosafot 0.512. One line per generation — which checkpoint to use and which are…

Model · Image and text to text

Qwen3.6-REAP-18B-A3B-onw

Ryu Gyo

onw（俺のNPUがこんなに動くわけない）用に変換したモデルです。onw は LLM を Intel NPU だけで動かすエンジンで、このリポジトリにはモデルだけが入っています（エンジンは別にダウンロードします）。 Qwen/Qwen3.6-35B-A3B の エキスパートを半分に刈り込んだ 版（各層 256 → 128 個、REAP 方式）。総パラメータ約 18B、1 トークンあたり約 3B が動きます。テキストと画像に対応。刈り込みなしの版は ryugyosoft/Qwen3.6-35B-A3B-onw。 私たちの知る限り、MoE と Gated DeltaNet を組み合わせたモデルを画像入力まで含めて NPU だけで動かす公開実装は onw が初めてです（2026 年 9 月、Fable 5.1 調べ）。16 GB の PC 向けに刈り込んだのがこの版です。 1. onw（エンジン）を入れます。 まだの場合は、onw の「インストール」 のとおり 1 行貼るだけです。Windows はスタートメニューで「PowerShell」を開き、次の 1 行を貼って Enter（Ubuntu は onw の README の curl... | bash）。 2. タスクトレイの onw アイコンから開く onw ウィンドウ の「モデル」タブで Qwen3.6-REAP-18B-A3B を選び、「ダウンロード」を押します（9.6 GB）。 3. 「サーバー」タブでモデルを選んで「ロード」を押します。初回だけ NPU 向けの準備に約 7 分かかります。 4.…

A 137M bi-encoder that ranks the functions of a repository by how likely they need to change to resolve a GitHub issue. It is nomic-ai/CodeRankEmbed (MIT) fine-tuned with an independent re-implementation of the SweRank data recipe (arXiv 2505.07849) on permissively licensed repositories only. No SweRank code or weights were used for training, filtering or mining. Licence: Apache-2.0. v0.2 (this revision, tag v0.2) was trained on Python and on Go, TypeScript, JavaScript, Rust, Java, PHP, C and C++ issues. v0.1 (Python only, the first release) stays available under the tag v0.1. The model's remote code (from CodeRankEmbed) reads the weights of the default branch even when revision= is given…

Model · Text to speech

Qwen3-TTS-12Hz-1.7B-CustomVoice-OpenVINO

X Girls

A ready-to-use OpenVINO IR export of Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice, prepared for Fern. The export preserves the upstream model weights and architecture; it is not fine-tuned. Generate speech with the model’s built-in speaker choices and supported style instructions. This variant does not include the Base model’s reference-audio cloning mode. Install the matching Qwen3-TTS model from Fern’s audio workspace. Fern downloads this ready-made export; users do not need to convert the source checkpoint. The first load compiles the IR for the selected device and can take longer than later loads. The following uses Fern’s Qwen3-TTS helper. Run it in an environment with the Fern…

Model · Text generation

StandardOne-3B-LoRA

Standard Thinking

Standard One scores a bounded set of answers for a supplied scenario and returns probabilities through POST /v1/systemone. It does not generate free-form response text. This repository contains the 3B LoRA adapter and a merge recipe; serving requires a merged checkpoint and the server code. In the reported served evaluations, 3B has a lower median latency on the measured short-request profile; 8B scores higher on the public standard and hard tiers. See Benchmarks for the measurement conditions and limitations. The figure combines results from different measurement paths. See Benchmarks for served versus offline conditions; measured 24–26 September 2026. - Send a state and a bounded rubric…

Model · Image and text to text

TerminalHorizon-122B-A10B

Zoushun

Scaling Agentic Data for Long-Horizon Terminal Intelligence TerminalHorizon is a fully automated data synthesis engine for long-horizon terminal agents. Grounded in real-world professional work, it constructs executable environments and generates training trajectories that connect agentic behavioral patterns across stages toward a shared goal. This repository releases TerminalHorizon-122B-A10B, obtained by supervised fine-tuning Qwen3.5-122B-A10B on The terminal is a shared interaction interface, not a restriction to software tasks. The training environments span 25 professional domains and 463 subdomains, including software engineering, machine learning, education, agriculture, healthcare…

Model · Image and text to text

TerminalHorizon-27B

Zoushun

Scaling Agentic Data for Long-Horizon Terminal Intelligence TerminalHorizon is a fully automated data synthesis engine for long-horizon terminal agents. Grounded in real-world professional work, it constructs executable environments and generates training trajectories that connect agentic behavioral patterns across stages toward a shared goal. This repository releases TerminalHorizon-27B, obtained by supervised fine-tuning Qwen3.5-27B on The terminal is a shared interaction interface, not a restriction to software tasks. The training environments span 25 professional domains and 463 subdomains, including software engineering, machine learning, education, agriculture, healthcare, finance…

Action Chunking with Transformers (ACT) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates. This policy has been trained and pushed to the Hub using LeRobot. Learn how to train and run it in the LeRobot act guide, or browse the full documentation. The policy consumes these observation features and produces these action features. Inputs Outputs New to LeRobot? These guides cover the full workflow: - Install LeRobot — set up the lerobot package. - Hardware setup — assemble, wire, and calibrate your robot and cameras. - Record data & train a policy — the end-to-end imitation-learning…

A fine-tune of Cloudflare/clef-flash (Qwen3.5-9B text decoder + 122M joint-schema head). It is distilled from the 27B Cloudflare/clef teacher and Claude Opus 5.5 author labels on synthetic decision data, with a KL anchor to the original clef-flash so general knowledge is kept. On the same 32 public Decision Index benchmarks it scores above both its base model and TypeSafe Jev 1.13. Release layout (same as Clef-flash, so the bundled loader works unchanged): merged bf16 backbone model-0000-of-00004.safetensors + fine-tuned fp32 head jointhead.safetensors + configs, tokenizer and jointschemamodel.py. The un-merged PEFT adapter (r = 32, all decoder Linear) is in adapter/, with its head and…

This gemma4 model was trained 2x faster with Unsloth and Huggingface's TRL library.

static quants of https://huggingface.co/singam96/LionCAD-Instruct For a convenient overview and download list, visit our model page for this model. weighted/imatrix quants seem not to be available (by me) at this time. If they do not show up a week or so after the static ones, I have probably not planned for them. Feel free to request them by opening a Community Discussion. If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files. (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants) Here is a handy graph by ikawrakow comparing some lower-quality quant…

Model · Text to speech

Qwen3-TTS-12Hz-0.6B-CustomVoice-OpenVINO

X Girls

A ready-to-use OpenVINO IR export of Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice, prepared for Fern. The export preserves the upstream model weights and architecture; it is not fine-tuned. Generate speech with the model’s built-in speaker choices and supported style instructions. This variant does not include the Base model’s reference-audio cloning mode. Install the matching Qwen3-TTS model from Fern’s audio workspace. Fern downloads this ready-made export; users do not need to convert the source checkpoint. The first load compiles the IR for the selected device and can take longer than later loads. The following uses Fern’s Qwen3-TTS helper. Run it in an environment with the Fern…

Model · Text to speech

Qwen3-TTS-12Hz-1.7B-Base-OpenVINO

X Girls

A ready-to-use OpenVINO IR export of Qwen/Qwen3-TTS-12Hz-1.7B-Base, prepared for Fern. The export preserves the upstream model weights and architecture; it is not fine-tuned. Generate speech and clone a voice from reference audio, subject to the source model’s supported languages and input requirements. Install the matching Qwen3-TTS model from Fern’s audio workspace. Fern downloads this ready-made export; users do not need to convert the source checkpoint. The first load compiles the IR for the selected device and can take longer than later loads. The following uses Fern’s Qwen3-TTS helper. Run it in an environment with the Fern scripts/ttsruntime directory on PYTHONPATH and the runtime…

Part of the Qwen3-VL-8B Hebrew / Cairo Genizah fine-tune series (https://github.com/AIStream-Peelout/historical-document-analysis). Flagship: v21b step 1200. Status: not adopted. The second arm of the merger ablation; use v21b. Trained 2026-08-30 → 09-01 in two Colab sessions (steps 100–1200, then 1300–1800; never reached 2,000); warm start v18b step 700; data identical to v19a. (colab/genizahfocusv19c.ipynb). Best eval loss of the v19 generation: 0.544 at 1200, 0.531 at 1800. Finding: the full-weight merger cured the pages v19b broke but each checkpoint broke a different handful of pages instead, and medians only tied v19a. Together with v19b this fixed the series decision: capacity in the…

EgoSteer-3B-RealMan is the generalist policy from EgoSteer: A Full-Stack System Towards Steerable Dexterous Manipulation from Egocentric Videos. It is a world-model-enhanced Vision-Language-Action (VLA) policy built on a Qwen3-VL backbone with a flow-matching action expert, post-trained from EgoSteer-3B-Base on real-world demonstrations collected on the RealMan robot. This is a robot-ready dual-camera (head + chest) checkpoint, corresponding to EgoSteer-9.6K-DG in the paper. For the base model to fine-tune on your own data, see EgoSteer-3B-Base. Our full-stack system integrates EgoSmith (data pipeline), Robot Stack (deployment), and EgoSteer (policy) to learn from 9.6k hours of large-scale…

Model · Zero-shot classification

Vela-2.0-0.3B

vLLM Semantic Router

Open Foundation Routing Models Routing decisions. Safety checks. Precise text spans. Vela 2.0 brings routing questions and span-level decisions into one model interface. Supply your options, labels and rubrics at request time; ask about the request, its context and the answer together. The 0.3B is the family's compact multilingual encoder, with an 8,192-token input budget and Torch or ONNX inference on CPU or GPU. 1. Route with your own criteria. Choose a route, check a policy condition, assign a score or select multiple labels. 2. Locate the text behind a signal. Return spans of personal information and unsupported claims with character offsets and probabilities. 3. Ask across the whole…

EgoSteer-3B-Base is the base world-model-enhanced Vision-Language-Action (VLA) policy from EgoSteer: A Full-Stack System Towards Steerable Dexterous Manipulation from Egocentric Videos. It is built on a Qwen3-VL backbone with a flow-matching action expert and a DINOv3 latent future-prediction objective used only during training, and learns a unified human-to-robot action space from large-scale egocentric human videos. This is the pretrained base checkpoint, corresponding to EgoSteer-9.6K-PT in the paper. For a robot-ready generalist, see Our full-stack system integrates EgoSmith (data pipeline), Robot Stack (deployment), and EgoSteer (policy) to learn from 9.6k hours of large-scale…