#Hugging Face

AICloudflare has published a new model called clef-omni on Hugging Face, according to a post from Julien Chaumond, who owns the account. The post links to the model page but gives no further details about its size, capabilities, or benchmarks.

AIMistral releases Voxtral Mini 4B Realtime Arabic, a streaming speech-to-text model for Arabic dialects and Modern Standard Arabic under the Apache 2.0 License. The model has about 4.4 billion parameters, is fine-tuned from Voxtral-Mini-4B-Realtime-2602, and reaches an average 8.82% Character Error Rate across seven Arabic benchmarks at a 480 ms transcription delay. It can be run with vLLM or Transformers 5.2.0 or later.

AIAi2's AI Infrastructure team replaced its priority-based GPU cluster scheduler with a system using GPU time budgets, hierarchical fair-share allocation, and a time-slicing contract. The team says the change turns decisions about how much GPU time each research project receives into a transparent administrative budgeting process. Its clusters, which range from 88 to 1024 GPUs including H100, B200, and B300 units, serve about 150 researchers facing demand two to three times available capacity.

AIQwen has published Qwen-Image-2.1-Turbo on Hugging Face, an accelerated checkpoint of Qwen-Image-2.1 for text-to-image generation and image editing with 8 denoising steps. The checkpoint uses the same 7B visual generation architecture, loads directly with QwenImage21Pipeline in Diffusers, and includes its recommended sampling schedule. It defaults to CFG=1 and uses prefix KV caching to reuse text and reference-image context across steps.

AIThree OpenAI safety researchers, Tomek Korbak, Jasmine Wang, and Mikita Balesni, say they were fired and that their terminations are scaring remaining employees. OpenAI says an investigation found they violated policies on handling sensitive information and denies firing anyone for raising safety concerns, without specifying the breach.

AINVIDIA has released a deployment model for Agile One S SSD pickup, based on GR00T N1.7 checkpoint 58000 and using three cameras: ego, left wrist, and right wrist. The repository republishes ONNX graphs, external tensor files, and two existing TensorRT BF16 engines without retraining or re-export, and the original export reported a numerical warning that full FP32, node, and BF16 parity did not pass all tolerances. The files are not a certified robot deployment or safety qualification.

AIJetBrains released Mellum2.1, a 12B mixture-of-experts coding model with 2.5B active parameters under Apache 2.0, emphasizing agentic programming. Under high load, its inference throughput in tokens is nearly twice that of Qwen3.5-9B in JetBrains' comparison, and multi-token prediction (MTP) speeds single-request responses by about 1.6x. The model is available on Hugging Face for local or private-infrastructure deployment, with GGUF and vLLM MTP support announced for later.

AIGoodfire has launched monitors that read a model's internal signals during agent work instead of reviewing its written output. The monitors are available to Baseten customers, who can choose risks to watch and set automated responses. In Goodfire's tests on Kimi K3, monitoring about 1,500 sessions cost roughly $51 versus about $10,000 for a top-tier AI judge, while catching 94% of malicious hacking sessions.

AIHugging Face is asking users to try Paper Pages at to help keep scientific research open. The post provides no further technical details, figures, or feature specifics about the platform.

AIClément Delangue of Hugging Face points users to a Hugging Face Space from Pollen Robotics where they can create a dance for Reachy Mini. The post gives no further details about how the dance tool works.

AIA Hugging Face blog author used the ML-Intern agent in HuggingChat to build six small models by writing detailed prompts that specify datasets, base models, baselines, smoke tests, and spending limits. The projects include a citrus disease vision-language model, a Huggy character LoRA, a camera-angle LoRA, a doodle-to-object LoRA, a 0.8B prompt rewriter, and a 4-step distilled Agate model, with total compute cost of about USD 103. Each project's prompts and public models are linked from the post.

Why it matters: The author shows how prompt structure, baselines, smoke tests, and budget caps shape an agent-driven training workflow, with per-project costs given.

AIGoogle released EmbeddingGemma 2 on Hugging Face, and Paige Bailey called it a step toward open models for open science. The post links the model and cites earlier EmbeddingGemma-based projects, including medical, geographic, oncology, and PubMed embedding models.

AILiquid AI released two open-weight decision models, d1-3B and d1-omni-600M (experimental), built on its Liquid Foundation Models and available on Hugging Face. d1-3B scores 48.57 on the Decision Index 0.2.1, the highest among decision models under 10B parameters, and answers a question in 16 ms on an NVIDIA Jetson AGX Thor and under 50 ms on a Jetson Orin Nano. The models support text and images (d1-3B) or text with image or audio (d1-omni-600M).

AIPerplexity has published both pplx-embed-v2 models on Hugging Face, with the collection available at the linked URL. The post provides no further details on model sizes, benchmarks, or capabilities.

AINVIDIA reports that fine-tuned Nemotron models reached gold-medal level at both IOI 2026, scoring 535.4 out of 600, and IMO 2026, scoring 30 out of 42. The IOI run was a live, unofficial, unsupervised benchmark, while IMO proofs were graded by official IMO graders. The post also releases checkpoints, datasets, a new 200-problem benchmark, and inference pipelines on Hugging Face and NeMo-Skills.

Why it matters: The post traces how SFT, RL, and a generate-verify-refine loop turned Nemotron into gold-level specialists for IOI and IMO, with the training and inference details shared.

AIA Hugging Face Space lets users try EmbeddingGemma 2 directly in their browser using WebGPU. The post shares only a demo link, with no further details on model size, benchmarks, or pricing.

AIGoogle releases EmbeddingGemma 2, its first native multimodal embedding model, built on Gemma 4 under Apache 2.0. It embeds over 100 languages, code, images, audio, and video into one vector, with an 8,192-token context and four sizes from 270M to 740M parameters. Matryoshka output dimensions of 768, 512, 256, or 128 are supported, and the model is available in Sentence Transformers and LiteRT-LM, with a reported 14% gain on MTEB Code.

Why it matters: The release extends an embedding model to text, code, images, audio, and video in one vector, a useful option for retrieval systems that mix media types.

AIPerplexity says its pplx-decider-v1.1-27b model supports a 250k context window and accepts text and image input. The weights are available on Hugging Face, and the model can be accessed through the Perplexity Decisions API.

AIllama.cpp added day-0 support for EmbeddingGemma 2, with GGUF builds available on Hugging Face under ggml-org. The post links to the embeddinggemma-2-GGUF model page as the way to run it.

AIGoogle has made EmbeddingGemma 2 available to download on Hugging Face and Kaggle. Availability in the Gemini Enterprise Agent Platform Model Garden is coming soon.

AIGoogle has released the weights for EmbeddingGemma 2 on Hugging Face and Kaggle. The release includes out-of-the-box support for LiteRT, MediaPipe, LangChain, and LlamaIndex.

AIJulien Chaumond reposted Mistral's announcement of Mistral Large 4, a 1T-parameter natively multimodal model with 49B active parameters. Mistral says it is available via API today, with open weights scheduled for release at the end of October, and is working privately with cybersecurity partners.

Why it matters: The post lays out Mistral Large 4's scale, multimodal design, and availability timeline, which helps readers gauge the open-weights landscape outside China.

AITencent has open-sourced a new video model called Prism, according to the post. The poster says the previews look solid, with a Hugging Face link provided for access.

AINVIDIA has published Nemotron-Labs-3-Competitive-Coding on Hugging Face, a competitive-programming specialist model built on Nemotron-3-Ultra. The model is available in the NVIDIA-Nemotron-Labs-3-Competitive-Coding-550B-A55B-NVFP4 repository, indicating a 550B-parameter total size with 55B active parameters in NVFP4 format.

AILiquid AI has released d1-3B, a 3B parameter multimodal model post-trained to return calibrated, typed answers to yes/no, choice, and score questions in one forward pass. The source reports a Decision Index 0.2.1 score of 48.57, the highest among models under 10B in its table, and 8 ms per decision on an NVIDIA RTX 4090.

Why it matters: The source gives benchmark scores against named peer models and edge latency figures across several hardware targets, helping readers judge fit for on-device decision pipelines.

AIMicrosoft and Hugging Face released ThinkingBox, a benchmark that grades AI agents on the terminal backend state and side effects they leave behind rather than their final responses. Each of 507 stateful business tasks runs 20 times from a clean backend, and the post reports pass@1, pass@20, and observed 20/20 counts, plus cost per successful and per dependable task across 18 models. The harness and dataset are available on Hugging Face, with the OpenEnv interface for running evaluations.

Why it matters: The post shows why checking the database state, not tool calls or final replies, exposes agent failures, and gives a repeat-run method for judging reliability.

AIAi2 released AstaBrief 8B, an open-weights model that turns a research question and retrieved literature excerpts into a cited report, along with its training data. The model runs as Fast mode in Asta, averaging 51.1 seconds per report versus 178.5 seconds for Thinking mode, about 3.5x faster. The post also describes filtering synthetic training data by citation density and building DPO pairs judged by two models that agreed.

Why it matters: The post explains how supervised fine-tuning, preference data, and citation-density filtering were used to build a cited-report model, which is useful for teams training their own models.

AILiquid AI shared a Hugging Face guide to multi-harness reinforcement learning for coding agents, in which a proxy records the token ids and logprobs vLLM samples so training works without changing the harness. Per the quoted post, LFM2.5-2.6B rose from 42% to 54% after training across four harnesses at once, and imitation fine-tuning on 3,189 rollouts from Qwen3.8-27B plateaued at 47.5%, below both RL runs. The proxy, trainer, tasks, SFT data, training code and seven trained models are described as open.