Qwen3.8-Omni-Flash Challenges Gemini, Unity Enhances AI Agent Development, and Google DeepMind's Dream-RSI Advances AI Learning

Anna SpatarProduct Expert at Amplify · Draft assembled with AI research tooling, edited by Anna Spatar

Today's AI news highlights Qwen's competitive multimodal model, Unity's new plugins for AI agent development, and Google DeepMind's innovative Dream-RSI for agent improvement.

The AI landscape continues its rapid evolution, with significant advancements in model capabilities, developer tools, and agent learning methodologies. Today's updates showcase how innovation is driving both performance and efficiency across various sectors, from multimodal AI to game development and agent training.

🚀 Qwen3.8-Omni-Flash Undercuts Gemini Flash Pricing While Matching Multimodal Benchmarks

Qwen has introduced its new Qwen3.8-Omni-Flash model, a multimodal AI designed specifically for AI agents. This model demonstrates the ability to process audio and video concurrently and independently utilize tools for tasks such as editing vlogs, translating clips, or summarizing movies. Notably, Qwen3.8-Omni-Flash nearly matches Google's Gemini 3.8 Flash on audio-video benchmarks while offering a significantly lower API cost, as reported by The Decoder [11]. This development positions Qwen as a strong contender in the multimodal AI space, offering a cost-effective solution for developers.

🎮 Unity Launches Official Plugins for Claude Code and OpenAI Codex

Unity has released official plugins for Claude Code and OpenAI's Codex, aiming to streamline AI agent development within its platform. These plugins are designed to prevent AI agents from relying on outdated tutorials, ensuring developers have access to the most current and effective tools. According to The Decoder, this initiative by Unity will enhance the efficiency and accuracy of AI agent creation, providing developers with better resources for building intelligent systems [14]. This move underscores Unity's commitment to supporting the integration of advanced AI capabilities into game development and other interactive experiences.

🧠 Google DeepMind's Dream-RSI Helps AI Agents Improve by “Dreaming” About Past Attempts

Google and DeepMind have introduced Dream-RSI, an innovative method that allows AI agents to improve by simulating past search runs. This technique enables agents to test new strategies without the need for costly recalculations. In tests, Dream-RSI either matched or surpassed existing results, reducing the number of iterations required by up to 2.43 times, as detailed by The Decoder [18]. The core AI model remains unchanged, with only the search strategy adapting, highlighting an efficient approach to agent learning and optimization.

📈 Vals Aims to Become the Gold Standard for AI Benchmarking

Vals AI, a startup backed by Andreessen Horowitz, is working to establish a more neutral and trustworthy resource for AI benchmarking. In an environment increasingly saturated with various AI models, Vals hopes to provide a reliable standard for evaluating performance and capabilities. TechCrunch AI reports that this initiative seeks to bring clarity and consistency to the assessment of AI models, which is crucial for informed development and adoption across the industry [17].

The industry is clearly moving towards more sophisticated, efficient, and verifiable AI solutions across diverse applications.