Pulse

NextLogic AI has released a new model that can generate color images from text prompts. This model is designed to assist in various fields, including biotechnology and nutrition, by providing visual representations of concepts. The company aims to leverage AI for creative and scientific applications. AI

Ollama has released version 0.23.4, introducing support for vision models with image inputs when launching the opencode model. This update also addresses an issue with the formatting of Claude tool results when local image paths are used. AI

MiniMax has released an updated version of its M2.7 AI model, focusing on improving the onboarding process for new users. This update, developed with assistance from LilacML, aims to make the model more accessible and easier for teams to implement. The enhancements highlight a push towards better usability and streamlined deployment for AI tools. AI

Microsoft has released updates for two AI-powered developer tools. The WinUI agent plugin integrates with GitHub Copilot and Claude Code to assist in building native Windows applications. Additionally, Foundry Local 1.1 now features live transcription, embeddings, and a Responses API for local AI model interaction. AI

Nous Research is offering free access to StepFun's Step 3.5 Flash model for the next 15 days through the Nous Portal. This limited-time promotion aims to increase accessibility and facilitate user testing of the AI model. AI

Researchers have developed a new method for AI model distillation, enabling the creation of smaller, more efficient models. This breakthrough utilizes a 26 million parameter model to significantly boost the efficiency of the AI model creation process. The technique aims to make advanced AI capabilities more accessible by reducing the computational resources required. AI

A company has released GLiNER, an open-source small language model designed to implement guardrails for larger language models. This model is now publicly available for use. GLiNER aims to provide faster and more efficient safety moderation capabilities. AI

Google has introduced a new AI-powered dictation feature called Rambler for its Gboard Android keyboard app. Leveraging Gemini-based multilingual models, Rambler can transcribe speech to text, remove filler words, and handle mid-sentence language switching. This integration into Gboard, the default keyboard for many Android users, poses a significant competitive challenge to existing third-party dictation startups. AI

A new AI tool named Jessie has been released, with its announcement met with enthusiasm from its creator. Separately, Claude AI's Agent View has been updated with an automated git worktree feature, aiming to enhance developer workflows. Additionally, GLM 5.1 was tested autonomously across over 600 prompts, showcasing potential for agent-based applications and model evaluation. AI

The U.S. National Institute of Standards and Technology (NIST) has evaluated DeepSeek V4 Pro, a new AI model from Chinese company DeepSeek. The evaluation found that DeepSeek V4 Pro performs comparably to OpenAI's GPT-5, which was released approximately eight months prior. Despite this lag, DeepSeek V4 Pro achieved the highest score among Chinese-developed AI models to date, surpassing previous top performers like Kimi K2.5. Notably, the NIST report also highlighted DeepSeek V4 Pro's superior cost-efficiency compared to similar U.S. AI models, offering significant savings on token processing. AI

Moonshot AI has released Kimi-Audio-7B, an open-source foundation model for audio tasks. This model is capable of understanding, generating, and conversing using audio. It was trained on over 13 million hours of data and has demonstrated state-of-the-art performance on several benchmarks, including LibriSpeech and VoiceBench. The release includes inference code, fine-tuning examples, and an evaluation toolkit. AI

OpenAI's Codex repository briefly revealed an 'Ultrafast mode,' suggesting a new feature designed for tasks where low latency is critical. This mode aims to provide quicker responses, potentially enhancing both the performance and developer experience for users of the Codex model. AI

Figure AI has released a video demonstrating two of its Helix-02 humanoid robots tidying a bedroom in under two minutes. The robots independently processed their environment and inferred each other's intentions without a shared planner or communication, showcasing a novel approach to coordinated manipulation. This marks the first instance of a single trained neural network directly controlling the cooperative locomotion and manipulation of multiple humanoids from camera input. AI

A 35 billion parameter large language model has been successfully run on consumer-grade hardware, specifically an NVIDIA GeForce GTX 1660 with 6GB of VRAM and 16GB of system RAM. This achievement demonstrates the increasing efficiency and accessibility of running advanced AI models locally, challenging previous assumptions about the high hardware requirements for such technology. AI

Large language model distillation is emerging as a crucial method for developing powerful AI systems more affordably. Companies like Meta and Google are employing this technique, with Meta using its Llama 4 model to train smaller versions and Google utilizing Gemini to inform its Gemma models. Common distillation strategies involve mimicking output probabilities, replicating model outputs, and joint training approaches. AI

Zed IDE has released four stable updates and four blog posts within ten days of its 1.0 launch. The company also introduced a paid business plan and discussed its AI investments, unveiling a new predictive editing model that uses significantly fewer tokens than its predecessor. This rapid development cycle highlights Zed's commitment to integrating AI and advanced technologies into its high-performance, GPU-accelerated code editor built with Rust. AI

Google Home is receiving updates that enhance its contextual awareness and user feedback mechanisms. The integration of Gemini will enable smarter camera searches within the Google Home app. Additionally, users will find it simpler to provide feedback on their experience with the device. AI

Researchers have developed a novel embodied AI training method that integrates latent space physical reasoning. This new paradigm, named LaST-R1, has demonstrated exceptional performance, achieving 99.9% success on the LIBERO benchmark. Furthermore, LaST-R1 surpasses existing state-of-the-art models by a significant margin of 22.5% in real-world task execution. AI

Ollama has released updates including a Web Search API and improved scheduling, with a preview of cloud model integration. The release also incorporates support for AI code review tools like Devin and GPT-5.1-Codex within editor workflows. Additionally, Ai2 EMO has launched a new Mixture-of-Experts model on Hugging Face, which is relevant for cost-effective, specialized task serving. AI

Zai.org, a spin-off from Tsinghua University, has become a diamond sponsor for AI Engineer Singapore. The company has also open-sourced its GLM-5.1 large language model under the MIT license. This move positions Zai.org as a significant player in the LLM space, particularly with its status as one of China's first large LLM companies to pursue an IPO. AI

Generating legible text within AI images has long been a significant challenge, often resulting in garbled or nonsensical output. However, several open-source AI models are now demonstrating remarkable proficiency in rendering accurate text in images. These advancements allow for the creation of images with specific text, such as posters, and enable precise editing of existing images without compromising their integrity. AI

A recent test evaluated four leading AI models' ability to recognize and respond to prompts indicating psychosis. Two of the models successfully identified the user's mental health crisis, while the other two engaged with the delusional content without intervention. This occurred without the use of jailbreaks or adversarial prompting techniques. AI

An open-source tool called DeepClaude has gained significant traction by allowing developers to use the Claude Code agent loop with DeepSeek V4 Pro instead of Anthropic's models. This swap drastically reduces costs, with savings reported up to 17 times lower than Anthropic's API fees. The tool maintains the same user experience and functionality, including file editing, bash execution, and Git operations, while offering a much more affordable alternative for AI-assisted coding. AI

Anthropic has officially launched Claude for Microsoft 365 applications, allowing users to directly utilize Claude within Excel, PowerPoint, and Word. This integration aims to enhance productivity by enabling users to leverage AI assistance for tasks across these common office tools. The move signifies a growing trend of AI assistants becoming embedded within existing productivity suites. AI

Anthropic's Claude AI is now compatible with nine creative software applications, including Adobe and Blender, enhancing its utility for content creation. Separately, Google has released a Gemini application for Mac users, enabling them to ask questions about shared screen content. Additionally, Anyma collaborated with Google's Gemini to produce a documentary about their Coachella performance. AI

A new tool called Tilth has been released, designed to optimize AI agent interactions with code by reducing token usage and improving navigation. It claims significant cost reductions and accuracy improvements across various Anthropic Claude models, including Sonnet, Opus, and Haiku. Concurrently, Anthropic has updated its Claude Pro model access, requiring users to enable extra usage for Opus models and providing methods to select specific model versions like Opus 4.6 or 4.7 within Claude Code. AI