📦 Open Source

38 articles

🟢 📦 Open Source May 23, 2026 · 4 min read

Kedro: version 1.2.0 brings the @experimental decorator and a LangGraph agentic starter for GenAI pipelines

Linux Foundation AI project Kedro released version 1.2.0 along with Kedro-Viz 12.3.0. The new @experimental decorator enables marking APIs under development, and the starter project support-agent-langgraph demonstrates integration with LangGraph orchestration and Langfuse/Opik prompt management tools. Kedro-Viz gains Mermaid diagrams and node preview extensibility for improved pipeline debugging.

🟡 📦 Open Source May 21, 2026 · 2 min read

Stability AI: Stable Audio 3.0 with open-weight models and generation up to 6 minutes

Stability AI released Stable Audio 3.0 on 20 May 2026 — a family of 4 audio models (Small SFX, Small, Medium, Large) of which three are open-weight and available on Hugging Face. The key advance is generation of audio up to 6 minutes (the previous version produced 47 seconds), along with support for audio inpainting, causal continuation, and LoRA fine-tuning. All models are claimed to have been trained exclusively on licensed data.

🟢 📦 Open Source May 20, 2026 · 2 min read

LangChain: The agent that fixes agents — how LangSmith Engine was built

LangChain published a detailed technical overview of LangSmith Engine — an autonomous agent that analyzes errors in production AI agents and proposes concrete fixes. It compresses thousands of traces, classifies them with a screener sub-agent, and generates validated evaluators for the Issue Board.

🟡 📦 Open Source May 19, 2026 · 2 min read

PyTorch: ExecuTorch MLX Delegate delivers 3–6× faster model execution on Apple Silicon GPUs

The PyTorch team released the experimental ExecuTorch MLX Delegate — a backend that leverages the Apple MLX framework and Metal GPU kernels for 3 to 6 times greater throughput on Apple Silicon chips. Supports Llama 3.2, Qwen 3, Phi-4 mini, Whisper and Voxtral real-time streaming transcription.

🟢 📦 Open Source May 19, 2026 · 2 min read

GitHub: Copilot Spaces API now generally available

GitHub announced the general availability of the REST API for Copilot Spaces, allowing teams to programmatically create, configure and delete contextual AI workspaces. The new interface is especially useful for organizations managing large numbers of Spaces without relying on manual workflows.

🟡 📦 Open Source May 18, 2026 · 3 min read

arXiv:2605.15041 CAST Framework: Case-Based Calibration for LLM Tool Use Achieves +5.85pp BFCLv2 and -26% Reasoning Length

CAST is a new arXiv paper published on May 14, 2026, by Renning Pang, Tian Lan, Leyuan Liu, Piao Tong, Sheng Cao, and Xiaosong Zhang, introducing a case-based calibration framework for LLM tool use. The approach treats historical execution trajectories as structured information for reinforcement learning — achieving up to +5.85 percentage points execution accuracy improvement over the BFCLv2 baseline and a 26% reduction in average reasoning length.

🟢 📦 Open Source May 18, 2026 · 5 min read

arXiv:2605.15706 Differentiable Mixture-of-Agents: dynamic per-step agent routing achieves SOTA across 9 benchmarks

Differentiable Mixture-of-Agents is a new arXiv paper published on May 15, 2026 by Xingjian Wu, Junkai Lu, Siyu Yan, Xiangfei Qiu, Jilin Hu, Chenjuan Guo, and Bin Yang that introduces a differentiable routing mechanism for multi-agent LLM collaboration. The system dynamically selects and activates agents per reasoning step instead of using fixed topologies, achieves SOTA results across 9 benchmarks, and adapts at test-time without external annotations via predictive entropy self-supervision.

🟢 📦 Open Source May 18, 2026 · 4 min read

arXiv:2605.15100 Dual-Dimensional Consistency: 10× Token Consumption Reduction with Maintained Accuracy Across Five Benchmarks

Dual-Dimensional Consistency is a new arXiv paper published on May 14, 2026, by Rongman Xu, Yifei Li, Tianzhe Zhao, Yanrui Wu, Bo Li, and Hang Yan addressing inference-time scaling efficiency. The framework combines a Confidence-Weighted Bayesian protocol and Trend-Aware Stratified Pruning — across five benchmarks it demonstrates over 10× reduction in token consumption while maintaining or improving accuracy over strong baselines.

🟢 📦 Open Source May 18, 2026 · 3 min read

arXiv:2605.15015 Small Private LM: Competitive Results in Educational Assessment Design with Human-in-the-Loop Recommendations

Small, Private Language Models as Teammates for Educational Assessment Design is a new arXiv paper published on May 14, 2026, by Chris Davis Jaldi, Anmol Saini, Shan Zhang, Noah Schroeder, Cogan Shimizu, and Eleni Ilkou. A systematic comparison of smaller models against larger alternatives in generating pedagogically aligned assessment questions — smaller models reach competitive results with privacy benefits, but the authors emphasize that model-based evaluations show systematic inconsistencies and recommend a Human-in-the-Loop approach.

🟢 📦 Open Source May 18, 2026 · 3 min read

arXiv:2605.15109 Traversal Context: Agentic GraphRAG Must Document Visited-but-Uncited Entities for True Provenance

Why Neighborhoods Matter is a new arXiv paper published on May 14, 2026, by Riccardo Terrenzi, Maximilian von Zastrow, and Serkan Ayvaz (accepted for the IJCAI-ECAI 2026 Joint Workshop on GENAIK and NORA). The authors argue that agentic GraphRAG systems must treat citation faithfulness as a trajectory-level problem — true provenance covers not only cited evidence, but also visited-but-uncited entities that influence model reasoning.

🟢 📦 Open Source May 18, 2026 · 3 min read

xAI SDK Python v1.13.0: prepare_extension() Enables Batch Video Extension for Generated Clip Series

xAI SDK Python v1.13.0 was released on May 16, 2026 (commit author @double-di, PR #141) and introduces the new prepare_extension() method for batch video extension. The function extends the video API introduced in v1.10.0 by adding batch processing capabilities — developers can now prepare extensions for a series of clips in a single call instead of sequentially for each individual clip.

🟢 📦 Open Source May 16, 2026 · 3 min read

arXiv:2605.15132 APWA: distributed architecture for parallel agent workflows — non-interfering subproblems without cross-communication

APWA Distributed Architecture for Parallelizable Agentic Workflows is a new multi-agent system architecture paper published May 15, 2026 on arXiv by Evan Rose, Tushin Mallick, Matthew D. Laws, Cristina Nita-Rotaru and Alina Oprea. The system decomposes agentic workflows into non-interfering subproblems executed on independent resources without cross-communication. APWA scales on tasks where prior systems fail completely.

🟢 📦 Open Source May 16, 2026 · 4 min read

arXiv:2605.14968 GraphFlow: clinical pilot 97.08 % completion rate through formally verifiable visual workflows

GraphFlow is a new visual workflow system for reliable agentic AI published on May 15, 2026, on arXiv by Drewry H. Morris V, Luis Valles, and Reza Hosseini Ghomi of MedFlow Inc. The system addresses the compounding error problem (a 10-step process with 90 % per-step reliability succeeds only 35 % of the time) through a formally verifiable diagram-as-specification approach. A one-year clinical pilot across three sites executed 8,728 workflow runs with a 97.08 % completion rate using an early prototype.

🟡 📦 Open Source May 15, 2026 · 3 min read

arXiv:2605.15040 Orchard: open-source agentic framework achieves 67.5% on SWE-bench Verified with three specialized recipes

Orchard is a new open-source agentic modeling framework published May 14, 2026 on arXiv (Baolin Peng, Wenlin Yao, and 12 co-authors). The framework combines a lightweight environment layer with three specialized training recipes — SWE (software engineering), GUI (vision-language), and Claw (personal assistants). The Orchard-SWE variant achieves 67.5% on SWE-bench Verified after RL training, making it the state-of-the-art open-source solution for coding agents.

🟡 📦 Open Source May 15, 2026 · 3 min read

LangChain: Labs Research Program for Autonomous Agents — Partners Harvey, NVIDIA, Prime Intellect, Fireworks, and Baseten

LangChain Labs is a new applied research program announced on May 14, 2026, by Harrison Chase, targeting autonomous agent improvement from operational data — production traces, user feedback, and evaluation results. LangSmith serves as the backbone for collecting trace signals. Initial partners include Harvey (legal AI), NVIDIA (GPU/infra), Prime Intellect (distributed compute), Fireworks (inference), and Baseten (deployment).

🟡 📦 Open Source May 14, 2026 · 2 min read

LangChain: LangSmith Engine automates agent debugging — error clustering, root cause, PR and evaluator recommendations

LangSmith Engine is a new LangChain platform published on May 13, 2026, that automates the production feedback loop for AI agents. The engine analyzes trace data from production runs, clusters errors by pattern, diagnoses root cause in code, and generates concrete pull request suggestions plus evaluator definitions. The goal is a shift from manual debugging toward continuous automated improvement.

🟡 📦 Open Source May 14, 2026 · 2 min read

PyTorch: Version 2.12 brings device-agnostic torch.accelerator.Graph, MX quantization and 100× faster linalg.eigh

PyTorch 2.12 is the new production release of the PyTorch framework published on May 13, 2026, with 2,926 commits and 457 contributors. Key features: torch.accelerator.Graph device-agnostic API for CUDA, XPU and out-of-tree backends, torch.export support for Microscaling MX quantization (MXFP4/6/8), linalg.eigh up to 100× faster on CUDA via cuSolver, and torch.cond inside CUDA Graphs. TorchScript has been formally removed.

🟡 📦 Open Source May 13, 2026 · 2 min read

LangChain: Delta Channels in LangGraph reduce long-running agent storage 41× via incremental checkpoints

LangGraph Delta Channels is a new LangChain state-update mechanism released on May 12, 2026, that solves O(N²) storage explosion in long-running agents. Instead of a full snapshot at every step, Delta Channels record incremental changes and take a periodic snapshot every 50 steps. A benchmark workload shows 41× storage reduction; the update ships in Deep Agents v0.6 and LangGraph v1.2.

🟡 📦 Open Source May 13, 2026 · 2 min read

PyTorch: ExecuTorch comes to Arm Cortex-A, Cortex-M and Ethos-U85 NPU for edge AI inference

ExecuTorch on Arm is a new PyTorch Foundation initiative published on May 12, 2026, that extends the ExecuTorch runtime to Arm Cortex-A and Cortex-M CPUs and Ethos-U NPU accelerators. The OPT-125M transformer and MobileNetV2 model run on Raspberry Pi 5 and Ethos-U85 with 256 MAC units, and the Arm Education repository brings hands-on labs for edge AI deployment.

🟡 📦 Open Source May 7, 2026 · 2 min read

AMD: vLLM-ATOM plugin brings Instinct optimisations without changing vLLM code

AMD has released vLLM-ATOM, an open-source plugin that integrates Instinct GPU optimisations into the vLLM production framework without any changes to the upstream source code. It activates automatically through Python entry_points, supports dense and MoE models including Kimi-K2.5 and DeepSeek V3/R1, and uses AITER kernels for fused MoE and flash attention.

🔴 📦 Open Source May 6, 2026 · 2 min read

Allen Institute: MolmoAct 2 is the first open-source robotics foundation model to outperform GPT-5 and Gemini 2.5 Pro

MolmoAct 2 is an open-source robotics foundation model released on May 5 by Allen Institute for AI. The model achieves 63.8/100 on embodied-reasoning benchmarks, outperforms GPT-5 and Gemini 2.5 Pro, accelerates inference 37×, and is the first base model with built-in bimanual capabilities.

🟢 📦 Open Source April 30, 2026 · 3 min read

IBM Granite 4.1: open-source family of 3B/8B/30B Apache 2.0 models trained on 15T tokens shows that a dense 8B model matches 32B MoE

On April 29, 2026, IBM published on the HuggingFace blog the details of building the Granite 4.1 model family — 3B, 8B, and 30B dense variants under the Apache 2.0 license. Trained on ~15T tokens through a 5-phase pipeline strategy, with a 4-phase RL using GRPO+DAPO loss. Granite 4.1-8B Instruct matches or surpasses the previous Granite 4.0-H-Small (32B-A9B MoE) on most benchmarks — showing that dense models reach MoE quality at the same activation budget.

🟡 📦 Open Source April 29, 2026 · 2 min read

Marco-MoE: Open-Source Multilingual MoE with 5% Active Parameters Outperforms Dense Models with 3-14× More Activations

Marco-MoE is a new open-source family of sparse Mixture-of-Experts models published on April 28, 2026, by a team led by Jiang, Zhao, and colleagues. The models activate only about 5% of total parameters per token, are trained via upcycling from dense models on 5 trillion tokens, and the Instruct variants outperform dense competitors with 3 to 14 times more activated parameters. Weights, dataset, and training recipe are publicly released.

🟡 📦 Open Source April 28, 2026 · 4 min read

OpenAI releases Privacy Filter: 1.5B parameters, Apache 2.0 license, 128K context, and state-of-the-art detection of eight PII categories in a single pass

OpenAI has released Privacy Filter — an open-source personally identifiable information detector with 1.5 billion parameters (50M active), a 128,000-token context, and an Apache 2.0 license. It detects eight PII categories in a single pass and achieves state-of-the-art results on the PII-Masking-300k benchmark, with multilingual support.

🟢 📦 Open Source April 28, 2026 · 4 min read

ONNX v1.21.0 releases with Opset 26: new CumProd and BitCast operators, 2-bit type support, and Python 3.14 free-threading experiment

On April 27, 2026, the Linux Foundation AI & Data Foundation released ONNX v1.21.0 — introducing Opset 26 with the CumProd and BitCast operators, 2-bit type support, experimental Python 3.14 free-threading, and improvements to integer division consistency and compiler security.

🟡 📦 Open Source April 24, 2026 · 3 min read

Google DeepMind Decoupled DiLoCo: 20× lower network bandwidth for AI training across geographically distributed datacenters

Google DeepMind has introduced Decoupled DiLoCo, a distributed architecture for training AI models. It reduces the required network bandwidth from 198 Gbps to 0.84 Gbps across 8 datacenters and achieves 88% goodput compared to 27% with conventional methods.

🟡 📦 Open Source April 24, 2026 · 3 min read

vLLM introduces DeepSeek V4 with 8.7× smaller KV cache: one million token context on standard GPU hardware

vLLM published full integration of V4-Pro and V4-Flash models on the same day as DeepSeek's release, with an 8.7× smaller KV cache compared to V3.2 at one million token context. The combination of sparse attention and aggressive compression enables serving on standard GPU hardware.

🟢 📦 Open Source April 24, 2026 · 2 min read

Allen AI: OlmoEarth embeddings enable landscape segmentation with just 60 pixels and F1 score of 0.84

Allen Institute for AI has launched OlmoEarth Studio with three model sizes (Nano, Tiny, Base) for satellite embeddings. The models achieve an F1 score of 0.84 for landscape segmentation with only 60 labeled pixels and support change detection and PCA visualization.

🟡 📦 Open Source April 23, 2026 · 2 min read

Apple at ICLR 2026 introduces ParaRNN: parallel training of nonlinear RNNs with 665× speedup

Apple presented five research papers at ICLR 2026 in Rio de Janeiro, with the most notable being ParaRNN — a method enabling parallel training of nonlinear recurrent neural networks with a 665× speedup over sequential approaches, scaling RNNs to billions of parameters to compete with transformers.

🟡 📦 Open Source April 23, 2026 · 2 min read

Linux Foundation publishes RGAF guide with 35 open-source tools for responsible AI

Linux Foundation AI & Data published a practical guide for implementing the RGAF (Responsible Generative AI Framework) across nine dimensions of responsible AI, with a catalog of 35 concrete open-source tools and alignment with NIST AI RMF, EU AI Act, ISO/IEC 42001, and OECD principles.

🟢 📦 Open Source April 22, 2026 · 2 min read

WorldDB: memory engine combining vector database and world graph for AI agents

WorldDB is a new memory engine for AI agents that combines a vector database with a graph of nested worlds and an ontology. It reconciles knowledge at write time and prevents contradictions in agent memory, achieving 96.40% accuracy on the LongMemEval-s benchmark.

🟡 📦 Open Source April 21, 2026 · 3 min read

Allen Institute BAR: Modular Post-Training with Mixture-of-Experts Delivers +7.8 Points for Math on OLMo 2 7B

BAR (Branch-Adapt-Route) is a new modular approach to post-training from the Allen Institute for AI that enables independent training of domain experts — math, code, tool use, safety — and their combination into a unified mixture-of-experts model. Results on OLMo 2 7B: 49.1 average score, +7.8 points for math and +4.7 for code over the baseline retraining.

🟡 📦 Open Source April 21, 2026 · 3 min read

AMD FLy: Training-Free Speculative Decoding Delivers 5.21× Speedup on Llama-3.3-405B with Over 99% Accuracy

AMD FLy is a new training-free speculative decoding method that achieves 4.80× to 5.21× speedup on Llama-3.3-405B and 2.74× on Llama-3.1-70B through semantic acceptance of draft tokens, with accuracy above 99%, requiring no additional model training.

🔴 📦 Open Source April 18, 2026 · 3 min read

Google Gemma 4: four open models, 31B dense ranks third on the Arena leaderboard, Apache 2.0 license

Gemma 4 is Google's new generation of open models in four variants: E2B for mobile devices, E4B for edge devices, 26B MoE with 3.8 billion active parameters, and 31B dense. The 31B holds third place on the Arena open-model leaderboard and reportedly outperforms models 20 times its size, while the 26B MoE ranks sixth. All models are multimodal (text, image, video, audio), support 140 languages, offer up to 256K token context, and are released under the Apache 2.0 license.

🟢 📦 Open Source April 17, 2026 · 2 min read

HuggingFace: guide to training multimodal embedding and reranker models

HuggingFace has published a detailed guide for fine-tuning multimodal embedding and reranker models through the Sentence Transformers library. The focus is on unifying text and images in a shared embedding space, enabling semantic search across heterogeneous data. The primary application is in RAG systems working with a mix of documents, tables, images, and scans.

🟢 📦 Open Source April 17, 2026 · 2 min read

HuggingFace: Transformers-to-MLX skill for Claude Code brings AI-assisted model porting to Apple Silicon

HuggingFace has published a 15,000-word Transformers-to-MLX skill that uses Claude Code for porting Transformers models to the MLX-LM platform for Apple Silicon. The skill includes a test harness that independently verifies results, eliminating the problem of LLM hallucinations, and addresses the growing challenge of open-source projects where AI agents increase pull request volume by 10 times.

🟢 📦 Open Source April 16, 2026 · 2 min read

LangChain: How We Made Our Documentation Test Itself Using Deep Agents

LangChain has built an automated documentation testing system using Deep Agents that prevents stale code examples in documentation. The system uses multiple skills that migrate inline code snippets into standalone test files, run them through GitHub Actions, and regenerate verified snippets, ensuring the documentation always reflects the actual state of the API.

🟡 📦 Open Source April 10, 2026 · 2 min read

Waypoint-1.5: AI generates interactive 720p worlds on an ordinary RTX 3090

Overworld has released Waypoint-1.5, a real-time video world model that generates interactive virtual environments at 720p/60fps on consumer GPUs such as the RTX 3090 and Apple Silicon Macs. The model was trained on ~100x more data than the previous version and is available open source on HuggingFace.