Global AI Native Industry Insights – 20260806 – Cursor | Black Forest Labs (FLUX) | Mistral AI | more
Today’s digest highlights significant developments from Black Forest Labs and Mistral AI, among others. The news collectively showcases advancements in AI capabilities, focusing particularly on multimodal and enhanced computational models. Cursor has open-sourced their Mixture-of-Kittens MoE Training Megakernel optimized for NVL72s, while Black Forest Labs introduces FLUX 3 Video, featuring native audio and multi-frame image-to-video capabilities with up to 1080p output support. Additionally, Mistral AI’s release of Shieldstral, a 3-billion parameter open-weights content safety model, marks a step forward in safe content generation. These innovations underscore an ongoing trend towards more sophisticated and capable AI systems. Discover more in Today’s Global AI Native Industry Insights.
1. Cursor Open-Sources Mixture-of-Kittens MoE Training Megakernel for NVL72s
Cursor has open-sourced Mixture-of-Kittens (MoK), its production Mixture-of-Experts training megakernel designed for NVIDIA NVL72 systems. MoK fuses all MoE communication and computation into a single, fully deterministic kernel, achieving up to 2.37x faster performance than the strongest public baselines. The kernel currently powers training across tens of thousands of GPUs at Cursor and delivered a 1.41x improvement in end-to-end training throughput over Cursor’s prior stack. MoK was developed to support the scaling of Composer, Cursor’s agentic coding model.
Read more: https://cursor.com/blog/mixture-of-kittens
Video Credit: NotebookLM
2. Black Forest Labs Launches FLUX 3 Video with Native Audio, Multi-Frame Image-to-Video, and Up to 1080p Output
Black Forest Labs has launched FLUX 3 Video, a new AI video generation model offering text-to-video, image-to-video with multiple input frames, video continuation, and native audio including multilingual dialogue. The model supports outputs of up to 20 seconds at 1080p native resolution and includes a Draft mode for rapid, lower-cost idea exploration. FLUX 3 Video is available via API and through third-party tools at launch. The company announced that 2K, 4K resolutions and open-weights versions are planned for future release.
Read more: https://x.com/bfl_ai/status/2084693239836401907
Video Credit: @bfl_ml on X
3. Mistral AI Releases Shieldstral, a 3B Open-Weights Multimodal Content Safety Model
Mistral AI has launched Shieldstral, a 3-billion-parameter open-weights multimodal safety classifier designed for content moderation that can be deployed on-device. The model accepts plain-language moderation policies at inference time and returns calibrated safety scores for both text and images without requiring retraining. Shieldstral outperforms models up to seven times its size on text safety benchmarks and sets a new state of the art on multimodal moderation, while running on a single 16GB NVIDIA GPU. It is released under the Apache 2.0 license, and a full technical report is available on arXiv.
Read more: https://mistral.ai/news/shieldstral/
Video Credit: NotebookLM
4. Meta Launches Muse Code Beta, a Terminal Coding Agent Powered by Muse Spark 1.2
Meta has introduced Muse Code (beta), a terminal-based coding agent designed for long-horizon software engineering tasks. The tool is powered by Meta’s new Muse Spark 1.2 model and is capable of planning, implementing, and validating complex, multi-file changes across large repositories. Muse Code uses persistent sub-agents to tackle difficult engineering problems more efficiently and with less human intervention. The product targets developers who need automated assistance with large-scale, multi-step software development workflows.
Read more: https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2
Video Credit: @AIatMeta on X
That’s all for today’s Global AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.