PaddleFormers 1.2 is released with support for training DeepSeek-V4 and 128K+ long-context training. The update adds Context Parallel, Packing, Document Mask Attention, and the Muon optimizer, plus ultra-fused mHC, CSA, and HCA operators, DeepEP/HybridEP communication, and lossless FP8 training with AutoSubbatch memory balancing. The project is presented as fully open-source and is available on GitHub.
Qwen released Qwen3-ForcedAligner-0.6B-hf, a Transformers-format forced aligner that predicts timestamps for arbitrary units within up to 5 minutes of speech in 11 languages. The model accepts transcripts from any ASR system, and the documentation shows it paired with Qwen3-ASR-0.6B and NVIDIA Parakeet CTC. Until it ships in an official Transformers release, users must install Transformers from source.
👏Great to see PP-OCRv6 land on @huggingface 👍Better accuracy, plus transformers & ONNX Runtime backends. Definitely worth a try！ #OCR #OpenSource #HuggingFace
Cohere and Cohere Labs released Cohere Transcribe Arabic, an open-source 2B-parameter Arabic automatic speech recognition model under Apache 2.0. It is optimized for Arabic, Arabic dialects, English, and Arabic-English code-switched speech, using a Conformer encoder-decoder architecture supported natively in Transformers. The model's average WER of 25.87 and CER of 11.80 on the Open Universal Arabic ASR Leaderboard, as of 07.07.2026, is reported in the source.
NVIDIA GEAR's ENPIRE gives eight Codex agents a fleet of robots, GPUs, and a token budget to solve physical tasks with minimal human oversight. The author reports tasks such as tying zip-ties, organizing fine pins, and installing GPUs, and a faster time-to-solution with eight parallel robots than with fewer. Safety uses a kinematic limit that resets a robot leaving its envelope, a torque-limited gripper, and a frozen reward function classifier. The team says everything will be open-sourced.
PaddleOCR 3.7 adds an ONNX Runtime inference backend, switchable with a single parameter and no code changes. The release supports CPU via OpenVINO and GPU via CUDA and TensorRT, and introduces PP-OCRv6 Tiny, Small, and Medium models that run up to 3.9× faster on CPU.
Z.ai has released GLM-5.2, its flagship model for long-horizon tasks, which it says substantially improves on GLM-5.1 and supports a 1M-token context. The model adds IndexShare, which cuts per-token FLOPs by 2.9× at 1M context, and is released under the MIT open-source license.
AIWhy it matters: The source gives benchmark tables against named rival models and deployment settings, useful for judging where GLM-5.2 sits among current flagship models.
Moonshot AI released Kimi K3 on Hugging Face as an open-weight, native multimodal agentic model with 2.8T total parameters and 104B activated parameters. It supports a 1-million-token context window and text and image input, with weights released under the Kimi K3 License. The model card reports benchmark results for coding, agentic, and vision tasks against several closed models, and recommends vLLM, SGLang, or TokenSpeed for inference.
AIWhy it matters: The release pairs open weights with a 2.8T-parameter MoE architecture and benchmark tables against several named closed models, useful for comparing frontier capability claims.
Some cool ggml-based work by @mudler_it recently - make sure to check it out.
Xiaomi MiMo has released MiMo Code V0.1, an open-source AI coding assistant for the terminal under the MIT license. It ships with MiMo V2.5, a multimodal model offered free for a limited time with a million-token context window. The tool automatically loads existing Claude Code skills, MCP servers and commands, and reuses API configuration, and it supports providers including Anthropic, OpenAI, DeepSeek, Kimi and GLM.
AIWhy it matters: The post specifies MiMo Code's Claude Code compatibility and MIT license, which bear directly on whether existing coding-agent setups can migrate without rework.
ByteDance open-sourced inference code and model weights for Bernini-Diffusers, a full video generation and editing pipeline with an MLLM-based semantic planner and a DiT-based renderer. The release bundles a Qwen2.5-VL planner and Wan2.2 diffusion components in one self-contained directory, and the source recommends it over the renderer-only Bernini-R for complex instruction following.
EvoQuality is a ByteDance vision-language model for no-reference image quality assessment that generates pseudo-ranking labels through pairwise majority voting and refines them with GRPO, requiring no human-annotated quality scores. On the paper's setting, it raised weighted-average PLCC from 0.615 to 0.770 and SRCC from 0.570 to 0.726 over its Qwen2.5-VL-7B backbone. The model is recommended for research and pre-production assessment, not as the sole criterion for high-stakes decisions.
Z.ai released SCAIL-2, an open-source model that animates a reference character from a driving video without skeleton maps or inpainting masks. It also supports character replacement, multi-character scenes, and animal-driving, with 512p and 704p resolutions and inputs whose height and width are both divisible by 32.
Xiaomi MiMo and the TileRT team say a 1T model exceeds 1,000 tps on a single standard 8-GPU node using general-purpose GPUs. The speedup comes from FP4 quantization and DFlash, a block-masked parallel speculative decoding method that accepts more tokens per verification, with TileRT tailoring its compiler and kernels to these techniques. Open weights for the FP4 + DFlash checkpoint are available on Hugging Face.
Xiaomi MiMo and TileRT released MiMo-V2.5-Pro-UltraSpeed, which the post says reaches output speeds above 1,000 tokens/s on a 1 trillion parameter MoE model. The post says this runs on a single standard 8-GPGPU node rather than wafer-scale or pure on-chip SRAM hardware. UltraSpeed access is application-based from Jun 8 to Jun 23 (PDT), and the UltraSpeed API costs 3x the standard price.
ByteDance has open-sourced the 1.3B-parameter weights of its Bernini Renderer (Bernini-R), available on Hugging Face as ByteDance/Bernini-R-1.3B-Diffusers. Fine-tuned from Wan2.1-1.3B, the model performs close to the 14B variant on simple tasks such as style transfer, subtitle or watermark removal, and local editing, but lags on complex tasks such as human generation. The release requires a CUDA GPU, with an H100 recommended for FlashAttention-3.
Xiaomi MiMo has released MiMo-V2.5-Pro-FP4-DFlash, the FP4 backbone behind MiMo-V2.5-Pro-UltraSpeed, with MXFP4 quantization applied only to the MoE experts and a BF16 DFlash drafter for block-diffusion speculative decoding. The backbone has 1.02T total and 42B active parameters, and the drafter proposes blocks of up to 8 tokens per forward pass. The release is supported in SGLang, with example launch commands provided.
Xiaomi and TileRT released MiMo-V2.5-Pro-UltraSpeed, reporting decode speeds above 1000 tokens/s on a 1-trillion-parameter model using a single standard 8-GPU node. The API is priced at 3x MiMo-V2.5-Pro and is available by application only from June 9 to June 23, 2026. The speedup relies on FP4 quantization of MoE Experts, DFlash speculative decoding with an average coding acceptance length of 6.30, and TileRT compute kernels.
AIWhy it matters: The post traces how FP4 quantization, DFlash speculative decoding, and TileRT kernels combine to reach 1000+ tokens/s on a single 8-GPU node, which is useful for teams weighing inference throughput.
NitroGen just won CVPR Best Paper Honorable Mention!! We are making strides towards general-purpose embodied agents that master not only the real world physics, but also all possible physics across a multiverse of simulations. It’s been 4 years since MineDojo, our first embodied agent in Minecraft, won NeurIPS Best Paper. Congrats to everyone on the team!!
Cohere and Cohere Labs released North Mini Code 1.0, an open-weights 30B-A3B mixture-of-experts model for code generation and agentic terminal tasks, under Apache 2.0. The model has 256K context and 64K max output, and is trained for tool use. Its benchmark table lists Terminal-Bench v2 at 36.0, SWE-Bench Verified at 67.6, and LiveCodeBench v6 at 70.3, below Qwen3.6 on several tasks.
AIWhy it matters: The card lists benchmark results against Qwen3.6, Gemma4, and other models, showing where North Mini Code trails on some coding and agentic tasks.
👏Surprise: free to use — give it a try！ Baidu CoBuddy is now live on @novita_labs — A powerful code-focused model for developers and AI agents 🚀 ✨131K context · 65K output · native tool calling
MiniMax has released MiniMax-M3, a native multimodal model with a 1M-token context window, roughly 428B total parameters, and about 23B activated parameters. The model introduces MiniMax Sparse Attention, which the source says delivers 9× prefill and 15× decode speedups over M2 at 1M context. M3 supports enabled, adaptive, and disabled reasoning modes through the thinking parameter, and weights are available on Hugging Face.
AIWhy it matters: The source gives concrete attention-efficiency figures and three reasoning modes, which helps readers judge long-context cost against deployment choices.
ByteDance has open-sourced the inference code and model weights of the Bernini Renderer (Bernini-R), a DiT-based renderer paired with an MLLM-based semantic planner for video generation and editing. A diffusers-format version, ByteDance/Bernini-R-Diffusers, bundles the Wan2.2 base components with the Bernini-R transformer weights for direct loading, and the framework requires a CUDA GPU with PyTorch 2.5.1+cu124.
PaddleOCR and ERNIE Image are now available as official Dify plugins, bringing document parsing and image generation into Dify's agent workflows. PaddleOCR, powered by PP-OCRv5, PP-StructureV3, and PaddleOCR-VL, turns images, scanned PDFs, and multilingual documents into structured data for chunking, vectorization, and RAG, with private or on-prem deployment supported. ERNIE Image offers free generation, a Turbo mode with 8-step inference, and an OpenAI-style API.
MiMo-V2.5 is now available in OpenCode — free for a limited time. 🎉
Congratulations to NabuOCR 🎉 Thanks Zack for trusting us and building something truly meaningful with PaddleOCR. From ancient cuneiform tablets to AI-powered text recognition — technology's journey across millennia is what makes it so inspiring.
Wow. It looks like the @XiaomiMiMo v2.5 model is insanely good value :O (Price for each prompt shown after each answer. Context includes >40k tool descriptions, system prompt, skills, etc.)
Start experimenting with Stable Audio 3.0: 🤗 Small + Medium → @huggingface https://huggingface.co/collections/stabilityai/stable-audio-3 ⚡ Large → Stability AI API or self-host with an enterprise license https://stability.ai/stable-audio
Stability AI released Stable Audio 3.0, a family of four audio models trained on fully licensed data. Three of them, Small SFX, Small and Medium, have open weights on Hugging Face, while Large is available through the Stability AI API and enterprise self-hosting. Outputs can be distributed and commercialized under the Stability AI Community License, and organizations with more than $1M in annual revenue can use the Enterprise License.
AIWhy it matters: The source specifies which models are open-weight, their licensing terms, and clip-length limits, which matters for anyone deciding whether to build on them.
PaddleOCR 3.5 now supports Hugging Face Transformers as an inference backend, letting users run PP-OCRv5 and PaddleOCR-VL 1.5 models directly within the Transformers ecosystem. Users can select it with engine="transformers" while keeping the same PaddleOCR pipeline, which the post says eases integration for RAG and Document AI applications.
This is pretty cool - write PyTorch and export and run with MLX on Apple silicon
Google is rolling out Gemini 3.5 Flash globally for consumers in the Gemini app and Search AI Mode. It is also available to developers through the Gemini API, Google Antigravity, and Google AI Studio, and to businesses on the Gemini Enterprise Agent Platform.
AIWhy it matters: The post shows where each Gemini 3.5 Flash access path goes, from consumer apps to developer and enterprise platforms, which helps readers pick the right entry point.
ChatGPT is previewing a personal finance experience for Pro users in the U.S., who can securely connect financial accounts and see where their money is going. Users can ask questions based on the information they choose to connect, and the author says this follows the similar health records connection feature.
🎉 Huge congratulations to our partner — and PaddleOCR OCEAN Alliance member — RAGFlow on reaching an incredible 80K GitHub stars! ⭐🚀 We’re proud to build together with the RAGFlow team through deep PaddleOCR integration and document AI collaboration. More to come! 🌊
🧫 Petri, our open-source interactive behavioral-evals tool for alignment testing, is now a freestanding project living with Meridian! Try it out, and consider contributing!
Amazon Quick's new desktop app connects to email, calendar, Slack, local files, and other apps to flag priority messages, summarize information, send communications, and create agents that handle tasks. The post says it becomes more personalized with use, and the author describes using it to treat their inbox more like an archive.
Test of Time LLM from some legends...pretty fun!
Mistral AI has released mistralai/Mistral-Medium-3.5-128B-EAGLE, an EAGLE draft model for speculative decoding with the 128B dense Mistral Medium 3.5. The companion model, which the source says replaces Mistral Medium 3.1 and Magistral in Le Chat and Devstral 2 in Vibe, has a 256k context window, handles text and image input with text output, and is served with vLLM or SGLang using three speculative tokens. The model is released under a Modified MIT License that allows commercial use with exceptions for companies with large revenue.
Xiaomi released and open-sourced MiMo-V2.5-Pro, a 1.02T-parameter Mixture-of-Experts model with 42B active parameters and a 1M-token context window. The company reports gains in agentic tasks, software engineering, and long-horizon work, including a Rust SysY compiler task finished in 4.3 hours across 672 tool calls. Weights and tokenizer are on Hugging Face, and API pricing is unchanged.
AIWhy it matters: The release pairs a 1.02T-parameter open-weight model with long-horizon agent results and token-efficiency claims, useful for judging its fit in coding and agent workflows.
Nemotron Hackathon Winners - S Korea https://x.com/i/broadcasts/1dGYljNNWYDKX