#Open-source ecosystem

AIMistral AI released Leanstral 1.5 on Hugging Face as an open-source code agent model for Lean 4 proof assistant tasks. The model uses 119B total parameters with 6.5B activated per token, a 256k context length, and accepts text and image input. The source gives setup paths through Mistral Vibe and a local vLLM server, with recommended settings of temperature 1.0 and reasoning effort set to high for complex prompts. The model is licensed under Apache 2.0.

AIJim Fan announces ASPIRE, a system where coding agents use multimodal sensory traces from simulation and real robots to run evolutionary search over control programs and add the results to a growing skills library. The post claims up to a roughly 10x reduction in transfer learning tokens for sim-to-real and single-arm to bimanual transfer, and says the full stack will be open-sourced.

AIJim Fan introduces ASPIRE, a system in which coding agents observe multimodal sensory traces and run evolutionary search over control programs to distill skills into a growing library. The post says ASPIRE shares know-how rather than pixels or weights across the sim-to-real gap, reducing transfer learning tokens by up to about 10x. The author also says the full stack will be open-sourced and provides a gallery of 150+ tasks and 90+ skills.

AIXiaomi MiMo's account celebrated growing developer adoption of its open-weights models, crediting Cline for building on MiMo. Cline's linked post announced a $9.99/month subscription offering 2-5x discounted access to GLM-5.2 and other open-weight models including DeepSeek, Kimi, MiniMax, MiMo, and Qwen, with a $1.99 promo for sign-ups via npm i -g cline.

AIUnlimited-OCR, Baidu's long-context OCR model, now runs in vLLM, with a recipe provided for developers to try it. The background post says it parses entire books in one pass using Reference Sliding Window Attention (R-SWA), which keeps the KV cache fixed during decoding, and claims 35% faster throughput than DeepSeek-OCR at 6K output tokens.

AIPaddleFormers 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.

AISebastian Raschka's tutorial shows how to build a fully local coding agent by pairing an open-weight LLM served through an inference runtime with an open-source harness that can read files, edit code, and run commands. He recommends Qwen-Code for Qwen3.6, citing Nvidia's Polar paper, which found Qwen models performed best in Qwen-Code. The Qwen3.6 35B-A3B model is about 22 GB to download and needs roughly 30–40 GB of RAM.

AIPaddlePaddle's PP-OCRv6 Tech Deep Dive Ep.4 benchmarks the OCR models across A100, V100, Intel Xeon CPU, and Apple M4 setups. PP-OCRv6_tiny processes an image in 0.13s on A100, while PP-OCRv6_tiny with OpenVINO runs 3.9x faster than PP-OCRv5_mobile on Intel CPU. The post recommends Medium for high-concurrency APIs, Small for CPU document systems, Tiny for mobile or embedded devices, and Medium or Small for multilingual business use.

AIPaddlePaddle announced that PP-OCRv6 is now available on Hugging Face, with better accuracy than prior versions. The release also adds transformers and ONNX Runtime backends in PaddleOCR 3.7.

AIPaddlePaddle introduced PP-OCRv6, a lightweight OCR architecture built on the LCNetV4 backbone, in the first episode of its tech deep dive series. The post says PP-OCRv6_medium reaches 86.2% detection Hmean and 83.2% recognition accuracy, surpassing PP-OCRv5_server while running faster. Three model specs—Tiny, Small, and Medium—target edge CPU devices, balanced deployment, and industrial high-accuracy pipelines.

AIThe 8th BAAI Conference opened an immersive zone featuring SoulAgent, embodied intelligence, FlagOS, and AI for Science. Hands-on demonstrations covered humanoid robot tasks, multi-chip systems, and AI-assisted cardiac diagnosis, neuroscience, and drug discovery.

AIAndrew Ng argues that Anthropic's restrictions on building competing LLMs and a U.S. Commerce Department license requirement for foreign nationals led Anthropic to disable Fable access worldwide. He says this shows governments and providers can quickly cut off access to frontier AI, which may push nations and businesses toward sovereignty efforts and open-source alternatives, though training frontier models remains difficult.

AIJim Fan introduces ENPIRE, which gives eight Codex agents a fleet of robots, GPUs, and a token budget to solve physical tasks autonomously. The post reports that the system can tie zip-ties, organize fine pins, and install GPUs, and that eight robots exploring in parallel improve faster than fewer. The team plans to open-source everything.

AIMistral states that this model and upcoming ones will be open-weight. The company argues that open weights are critical for customer confidence and for research and developer communities. It contends that systems reachable only through someone else's interface cannot be owned, inspected, audited, or improved, especially if data recording can no longer be turned off.

AIZed's 12-week Guild program, its first cohort run this spring, had 33 active contributors merge 148 pull requests into the open-source editor. The top contributor, feitreim, merged 23 PRs, including fixes for Vim mode screen flickering and terminal ANSI rendering, and won a trip to Rust Week in Utrecht. Zed plans to organize Cohort 2 work into tighter groups around specific parts of the codebase.

AIByteDance has released Sa2VA-LLaVA-1.5-7B on Hugging Face, a model built on LLaVA-1.5-7B with a SAM2 grounding encoder that performs dense image and video referring segmentation alongside open-ended chat. The checkpoint is self-contained and loads with trust_remote_code=True without extra packages, and it is positioned as a LISA-comparable baseline within the Sa2VA family. Reported results include 80.3 cIoU on RefCOCO val and 54.8 J&F on MeViS (val_u).

AITri Dao's post proposes loading SSM states, computing with them, and not storing them to make hybrid model decoding about 2x faster. The trick aims to ease the bottleneck from Gated-DeltaNet and Mamba states in hybrid models such as Qwen 3.5 and Nemotron Ultra, unlocking speculative decoding for SSMs.

AIGeorgi Gerganov highlighted locate-anything.cpp, a native C++/ggml inference implementation of NVIDIA's LocateAnything-3B for open-vocabulary object detection and visual grounding. The project, from the LocalAI team, runs without Python on CPU and GPU, with quantized GGUF weights published on Hugging Face.

AIPaddlePaddle has released PP-OCRv6, a new OCR model series in Tiny, Small, and Medium sizes at 1.5M, 7.7M, and 34.5M parameters. The models reportedly improve detection accuracy by 4.9% and recognition accuracy by 5.1% over PP-OCRv5, with up to 5.2× faster CPU inference via OpenVINO. The unified model supports 50 languages and new scenarios including PCB, CAD drawings, digital tubes, and dot-matrix text, under Apache 2.0.

AIAwni Hannun praised a WWDC video by @angeloskath explaining how to set up agentic AI to run locally with MLX in an approachable way. He said the demos work well now, which was impractical less than a year ago before M5 and recent gains in open-weights models.

AIMoonshot AI published Kimi-K2.7-Code, a coding-focused agentic model built on Kimi K2.6, with a 1T-parameter MoE architecture and 32B activated parameters. The model card reports about 30% fewer thinking tokens than K2.6 and benchmark results against GPT-5.5 and Claude Opus 4.8, with weights and code released under a Modified MIT License.

Why it matters: The model card gives benchmark comparisons against GPT-5.5 and Claude Opus 4.8 on coding and agentic tasks, useful for judging its position among current coding models.

AIXiaomi's MiMo team released MiMo Code, an MIT-licensed terminal coding agent built on OpenCode for long-horizon programming tasks. The design centers on three areas: Max Mode parallel sampling that generates five candidates per turn, Goal-based completion verification, and a memory system that checkpoints session state and rebuilds context. The article reports offline benchmark results and a double-blind A/B test with 1,213 pairs in which MiMo Code's win rate exceeded 65% beyond 200 execution steps.

Why it matters: The article explains how MiMo Code handles long-horizon coding through computation, checkpointed memory, and cross-session evolution, useful for judging design tradeoffs in coding agents.

AICognition introduced FrontierCode, a coding benchmark built with open-source maintainers that measures whether models produce code a maintainer would merge. On FrontierCode Diamond, the hardest 50 tasks, Claude Opus 4.8 scored 13.4%, GPT-5.5 scored 6.3%, and Gemini 3.1 Pro scored 4.7%. The authors report 81% fewer misclassification errors than SWE-Bench Pro, though this figure comes from their own analysis of agent trajectories.

Why it matters: The benchmark's blocker and rubric design shows how code quality can be measured beyond unit-test correctness, which matters for judging coding agents.

AIByteDance 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.

AIMaintainers and NVIDIA engineers improved multi-GPU performance in ggml, the low-level engine behind llama.cpp, yielding significant gains on RTX systems. The work also lays groundwork for hardware-agnostic tensor parallelism in ggml. Details are in a technical blog from NVIDIA RTX Spark.

AIMiniMax published MiniMax-M3-MXFP8, an MXFP8 quantized variant of its native multimodal M3 model with 1M context, about 428B total parameters and about 23B activated parameters. M3 adds MiniMax Sparse Attention, which the source says yields 9× prefill and 15× decode speedups over M2 at 1M context. The model supports three thinking modes (enabled, adaptive, disabled) via the thinking parameter and can be served with SGLang, vLLM, or Transformers.

Why it matters: The release pairs sparse attention for 1M-token contexts with reported prefill and decode speedups over M2, useful for judging long-context serving costs.

AIByteDance 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.

AICognition has announced Devin Desktop, the next generation of Windsurf, which makes the Agent Command Center the default IDE surface for managing local and cloud agents, PRs, and context. Spaces let related agents share context, and Agent Client Protocol (ACP) support lets any ACP-compatible agent run alongside Devin. The IDE remains fully backwards-compatible with Windsurf, including editor extensions, keybindings, LSPs, and terminal workflows.

AIMiniMax released M3, an open-weight model with a 1M-token context window, native image and video input, and desktop operation support. The post credits a new sparse attention architecture, MSA, for long-context gains, reporting over 9x prefilling and over 15x decoding speedups and 59.0% on SWE-Bench Pro. The API and MiniMax Code are available now, with the technical report and open weights promised within 10 days.

Why it matters: The post pairs a new sparse attention design with benchmark figures and a 1M-token context window, letting readers judge the architecture's practical effect on long-context work.

AIXiaomi describes an end-to-end inference optimization for the MiMo-V2.5 series, centered on Hybrid SWA, which it says cuts KVCache storage to roughly 1/7 of Full Attention. The post covers a dual KVCache pool design, SWA-aware prefix cache matching, the GCache distributed cache, and scheduling changes, and reports cache hit rates averaging 93% in server-side observations. It also covers prefill and decode optimizations, multimodal encoder improvements, and open-source contributions to SGLang.

Why it matters: The post explains how Hybrid SWA's theoretical KVCache savings were realized in production through dual pools, SWA-aware prefix caching, and tiered storage, giving concrete engineering patterns for long-context inference.

AIPaddlePaddle has released PaddleOCR-VL 1.6, which sets a new state-of-the-art score of 96.33% on OmniDocBench for text, formula, and table recognition. It ranks first on OmniDocBench v1.5 and Real5-OmniDocBench, with gains in table, classic text, rare character, seal, spotting, and chart recognition. The version is fully compatible with the v1.5 architecture, requiring no migration.

AIFunAudioLLM has released Fun-ASR-Nano-2512-hf, a Hugging Face Transformers-compatible version of its end-to-end speech recognition model, which supports Chinese, English, and Japanese. The Chinese coverage includes 7 dialect groups and 26 regional accents, and a separate Fun-ASR-MLT-Nano-2512 checkpoint handles 31-language recognition. Developers can run the model natively in Transformers 5.17.0 without custom model code or trust_remote_code=True.

AIYann LeCun points to Project Tapestry, hosted by thealliance.ai, as the answer to fears that no credible Western open frontier model will emerge. The post is a short pointer to the project, with no details on models, benchmarks, or availability.

AIShanghai AI Laboratory's Intern Large Models introduces Intern-S2-Preview, a 35B scientific multimodal foundation model, and says it matches the trillion-scale Intern-S1-Pro on core scientific tasks. The post says it is the first open-source model with material crystal structure generation and strong general capabilities, with shared-weight MTP plus KL loss improving acceptance rate and speed. It is already supported by vLLM and SGLang, with weights on Hugging Face and ModelScope.

AIOpenAI's GPT-5.5 scored 0.43% and Anthropic's Opus 4.7 scored 0.18% on ARC-AGI-3, a set of 135 novel environments, according to ARC Prize's replay analysis of 160 runs. The analysis found three recurring failure modes: models perceived local action effects but failed to build global rules, mapped unfamiliar games onto known ones, and sometimes beat a level without learning the underlying mechanic. ARC Prize is open-sourcing its analysis package.

AISoumith Chintala, a PyTorch co-creator, says GitHub and critical work are becoming incompatible. The post is brief and gives no specifics on which GitHub issues or what alternatives he has in mind.

AINVIDIA's developer account announced a livestream titled "Scaling NemoClaw: Roadmap, OpenClaw Collaboration, and Real-World Integration" as part of Nemotron Labs. The post provides no further details on the roadmap, collaboration terms, or integration specifics.

AIXiaomi's MiMo-V2.5 is a native omnimodal model that understands text, image, video, and audio within one architecture. It is a sparse MoE with 310B total and 15B activated parameters, and supports up to 1M tokens of context. The repository also notes a config.json and tokenizer_config.json update that users who downloaded before commit 4da2748 should re-pull.

Why it matters: The repository documents a 310B-parameter omnimodal MoE with a hybrid attention design, useful for comparing long-context efficiency against other open multimodal models.

AINVIDIA AI Developer announced a live "Ask the Experts" broadcast on running Gemma 4 on DGX Spark. The post provides only a broadcast link and no further technical details.

AITri Dao points to Wentao Guo's explanation of a mathematical rewrite of the MoE backward pass, which reduces activation memory and speeds up training, especially for fine-grained MoE. He also notes the work leverages new Blackwell features, including 2CTA MMA and CLC, to build fast MoE kernels.