Open source

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.

AIMoonshot 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.

Why 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.

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.

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.

AIByteDance 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.

AIEvoQuality 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.

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.

AIZ.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.

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.

AIXiaomi 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.

Why it matters: The post shows how FP4 quantization, DFlash speculative decoding, and TileRT kernels combine to reach over 1,000 tps on a standard GPU node.

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.

AIXiaomi 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.

AIXiaomi 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.

Why 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.

AICohere 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.

Why 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.

AIBaidu CoBuddy is now available for free on Novita AI, a code-focused model aimed at developers and AI agents. It offers a 131K context window, up to 65K output tokens, and native tool calling. The model is served through Novita's serverless API for high-throughput, low-latency inference.

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.

AIMiniMax 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.

Why it matters: The source gives concrete attention-efficiency figures and three reasoning modes, which helps readers judge long-context cost against deployment choices.

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.

AIXiaomi's MiMo-V2.5 is now available in OpenCode at no cost for a limited time. The OpenCode announcement lists a 1M context window and support for reasoning, text, and image inputs.

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.

AIReflection, an open-source AI lab, has signed a memorandum of understanding with the US Department of Energy to explore strategic collaborations under the Genesis Mission. Through the partnership, Reflection will provide open-weight models to the DOE's 17 National Laboratories, customizable on lab-specific scientific data and deployed into researcher workflows. The Genesis Mission focuses on advanced nuclear, fusion and grid modernization, quantum ecosystem growth, and national security AI.

AIStability AI has launched Stable Audio 3.0 for experimentation. The Small and Medium models are available on Hugging Face, while the Large model is accessible through the Stability AI API or self-hosting with an enterprise license.

AIStability 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.

Why 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.

AIPaddleOCR 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.

AIGoogle 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.

Why 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.

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.