OpenAI + Jony Ive, Claude 4 Released, Google Launches Veo 3, LMArena Raises $100M, and India’s Multilingual LLM
OpenAI's long rumored partnership with Jony Ive is official, Anthropic releases Claude 4, Google launches Veo 3, LLMArena raises $100 million from a16z, and Sarvam's LLM for India's languages
OpenAI Acquires Jony Ive’s io for $6.5B to Blend AI with Hardware
OpenAI announced their acquisition of io, the hardware startup founded by Apple design legend Jony Ive, for $6.5 billion on May 21, 2025.
The deal will integrate io’s design expertise with OpenAI’s AI research to create AI-driven physical devices. io’s team will merge with OpenAI’s San Francisco operations, with Ive overseeing design for both hardware and software, though he won’t join as an employee.
The partnership has been brewing for years, and first announced by us in September 2023.
This move signals OpenAI’s ambition to extend AI beyond software into consumer hardware, potentially rivaling Apple’s ecosystem and becoming more vertically integrated.
For consumers, this could mean intuitive AI-powered devices, blending Ive’s minimalist design with ChatGPT’s conversational prowess.
Anthropic Releases Claude Opus 4 & Sonnet 4, ARR Tops $3 Billion
Anthropic quietly rolled out their latest frontier models, Claude Opus 4 and Claude Sonnet 4, boasting superior reasoning and multimodal capabilities.
These models outperform OpenAI and Google’s Gemini in select benchmarks, and Anthropic’s ARR topped $3 billion, driven by enterprise demand.
This outperformance is important as they face a real threat in Google’s rapidly improving Gemini models, which are capable and significantly cheaper.
While performance is critical, it helps Anthropic that Claude is the only state-of-the-art option for both AWS and Google Cloud Platform customers. Gemini is only available for GCP, and OpenAI is the only state-of-the-art option on Azure.
Google Launches Veo 3 for Advanced Video Generation at I/O 2025
Google announced Veo 3, their next-gen video generation model, at Google I/O, alongside Gemini 2.5 Pro and an AI Ultra subscription ($249.99/month).
Here is what’s included in their AI Ultra subscription.
Veo 3 offers improved realism and motion coherence, competing with OpenAI’s Sora and Kling 2.0. AI Ultra subscribers gain early access to Veo 3, bundled with 30TB storage and YouTube Premium. See one of their videos for yourself.
Google is moving very quickly thanks to their TPU infrastructure advantage, as they aren’t beholden to Nvidia’s GPUs. Thanks to their advantages and execution, Google is seen as the far and away favorite for best AI model by end of 2025, according to Polymarket.
LMArena Raises $100M from a16z to Democratize AI Model Access
LMArena, a platform for comparing and deploying AI models, raised $100M from Andreessen Horowitz. The funding will expand LMArena’s open-source model hub, offering developers free access to test LLMs like Claude, Gemini, and Llama. The platform’s usage has surged 200% YoY, reflecting demand for transparent model evaluation.
Here is their current leaderboard, for reference. Notice how Claude Opus 4 is indeed at the top for web development.
LMArena’s growth lowers barriers to experimenting with state-of-the-art models, fostering innovation in niche applications. Their neutral stance contrasts with OpenAI and Google’s walled gardens, potentially reshaping how startups choose models.
This raise signals venture capital’s bet on open AI ecosystems, but they still face competition from players like Hugging Face.
Sarvam Launches Sarvam-M, India’s Multilingual LLM
Sarvam, an Indian AI startup, launched Sarvam-M, a multilingual LLM optimized for India’s 22 official languages.
Built for low-resource environments, Sarvam-M targets enterprises and public sector use, with early adoption by Indian telecoms and e-commerce. The model’s context window supports up to 128k tokens.
Sarvam-M addresses India’s linguistic diversity, offering a cost-effective alternative to global models for local businesses. Their focus on low-compute deployment could inspire similar efforts in other emerging markets.
However, scaling training data for 22 languages poses challenges, and Sarvam has only raised $53 million, which pales to OpenAI’s $58 billion and Anthropic’s $14 billion. Can Sarvam carve a niche in Asia’s AI race, or will they be overshadowed by the much larger players?