Perceptron Inc. launched its flagship video analysis AI model Mk1 on Monday, pricing it at $0.15 per million input tokens and $1.50 per million output tokens — roughly 80-90% less than competing models from Anthropic, OpenAI, and Google. The startup’s proprietary model targets enterprise video analysis applications including security monitoring, marketing content optimization, and behavioral analysis.
Breakthrough Pricing in Video AI Market
The two-year-old startup’s pricing undercuts major competitors significantly. According to Perceptron’s announcement, Mk1 costs substantially less than Anthropic’s Claude Sonnet 4.5, OpenAI’s GPT-5, and Google’s Gemini 3.1 Pro for equivalent video processing tasks.
Co-founder and CEO Armen Aghajanyan, formerly of Meta FAIR and Microsoft, led the 16-month development of what the company calls a “multi-modal recipe” built from the ground up. The model addresses complex physical world understanding including cause-and-effect relationships, object dynamics, and physics comprehension.
Enterprises can test the model through Perceptron’s public demo site, allowing potential customers to evaluate performance before integration.
Thinking Machines Previews Real-Time AI Interaction
Meanwhile, Thinking Machines announced a research preview of “interaction models” designed to move beyond traditional turn-based AI conversations. The startup, founded by former OpenAI CTO Mira Murati and researcher John Schulman, developed native multimodal systems treating interactivity as core architecture rather than external software.
The models demonstrate reduced latency and improved performance on third-party benchmarks by processing inputs while simultaneously generating responses. However, the technology remains in limited research preview with broader availability planned for coming months.
Thinking Machines positions these interaction models as essential for AI systems requiring natural human collaboration, moving beyond the current input-wait-output paradigm that dominates existing AI interfaces.
Sakana AI’s RL Conductor Orchestrates Multiple LLMs
Sakana AI researchers introduced the “RL Conductor,” a 7-billion parameter model trained via reinforcement learning to automatically coordinate multiple large language models. The research paper details how the system dynamically analyzes inputs, distributes tasks among worker LLMs including GPT, Claude, and Gemini models, and coordinates multi-agent responses.
RL Conductor achieves state-of-the-art results on reasoning and coding benchmarks while using fewer API calls than manually designed multi-agent pipelines. Co-author Yujin Tang told VentureBeat that hardcoded frameworks like LangChain “fall short because they are inherently rigid” when facing heterogeneous user demands in production environments.
The technology powers Fugu, Sakana AI’s commercial multi-agent orchestration service, addressing limitations in current agentic frameworks that break when query distributions shift in real-world applications.
Apple iOS 26.5 Adds Encrypted RCS Messaging
Apple released iOS 26.5 for iPhone 11 series and newer devices, introducing end-to-end encrypted RCS messaging between iPhones and Android devices. The feature requires carrier support and currently operates in beta testing phase.
The update includes a “Pride Luminance” wallpaper and “Suggested Places” in Maps offering location recommendations based on local trends and user activity. However, the anticipated new Siri remains absent, with analysts expecting its debut in iOS 27 at WWDC on June 8.
Google’s Android ecosystem president Sameer Samat called the RCS encryption “big news” for cross-platform messaging security, highlighting the industry significance of Apple’s adoption.
OpenAI-Musk Trial Enters Final Phase
The high-stakes trial between Elon Musk and OpenAI entered closing arguments Thursday, with audio livestreamed on YouTube. Musk’s 2024 lawsuit accuses OpenAI of abandoning its founding mission to develop AI for humanity’s benefit in favor of profit maximization.
Sam Altman testified Tuesday to refute Musk’s characterizations, following earlier testimony from Microsoft CEO Satya Nadella, former OpenAI chief scientist Ilya Sutskever, and other key figures. The trial could significantly impact OpenAI’s future structure and ChatGPT development.
Musk claims Altman and co-founder Greg Brockman deceived him into funding the company before abandoning original goals. OpenAI maintains the lawsuit lacks merit and reflects Musk’s competitive interests in AI development.
What This Means
These developments signal intensifying competition across AI model capabilities and pricing. Perceptron’s dramatic cost reduction for video analysis could democratize enterprise video AI adoption, while Thinking Machines’ interaction models suggest the next evolution beyond current chat interfaces.
Sakana’s orchestration approach addresses a critical infrastructure need as enterprises deploy multiple AI models simultaneously. The automated coordination could reduce development complexity and operational costs for companies building sophisticated AI applications.
Apple’s RCS adoption removes a longstanding barrier to secure cross-platform messaging, potentially influencing enterprise communication strategies. The OpenAI trial outcome may establish important precedents for AI company governance and mission accountability.
FAQ
What makes Perceptron Mk1 significantly cheaper than competitors?
Perceptron built its video analysis model from the ground up using a proprietary “multi-modal recipe” over 16 months, allowing more efficient processing architecture. The company prices Mk1 at $0.15 per million input tokens versus significantly higher rates from Anthropic, OpenAI, and Google.
How do Thinking Machines’ interaction models differ from current AI?
Traditional AI models use turn-based interaction where users input queries and wait for complete responses. Thinking Machines’ models process new inputs while simultaneously generating responses, creating more natural, fluid conversations similar to human interaction patterns.
What is the significance of Apple adding encrypted RCS messaging?
Encrypted RCS enables secure messaging between iPhone and Android users for the first time, eliminating the security gap that previously existed in cross-platform communications. This addresses a major enterprise concern about messaging security across different device ecosystems.
