Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 🤖 Meta enters humanoid robotics space

• 🤔 Claude 4 / Anthropic’s reasoning model

• 🧭 OpenAI roadmap update for GPT 4.5 and 5

• 🚢 HPE’s ships NVIDIA Grace Blackwell system

Let’s get right into it!

Sign Up | Advertise | Tools

BIG TECH

Meta enters humanoid robotics space

Meta is making a major investment in humanoid robotics, forming a new team within its Reality Labs division to develop AI, sensors, and software for robots, with an initial focus on household tasks. Rather than building its own branded robot, Meta aims to create a foundational platform for other companies, similar to how Google’s Android and Qualcomm’s chips supported the smartphone industry. Led by Marc Whitten, formerly of GM’s Cruise, the project will leverage Meta’s advancements in AI, hand tracking, and sensor technology. The company is in discussions with robotics firms like Unitree and Figure AI, while also building prototypes to refine its platform.

Meta executives believe their AI expertise and data from AR/VR devices could accelerate humanoid robot development, a field that still struggles with basic tasks like folding clothes and carrying objects. With plans to invest $65 billion this year in AI and robotics infrastructure, Meta seeks to position its Llama AI as the standard for humanoid development while also prioritizing robot safety. Though humanoid robots are still years away from mainstream adoption, Meta sees them as a long-term focus that could reshape the industry.

BIG TECH

Claude 4 / Anthropic’s reasoning model

Anthropic is taking a different approach to AI reasoning by developing a hybrid model that allows developers to control computational resources using a sliding scale, unlike OpenAI’s fixed “low,” “medium,” and “high” reasoning levels. This flexibility enables businesses to optimize cost, speed, and efficiency, positioning Anthropic as a strong competitor in the enterprise market. While OpenAI aims for broader consumer adoption with ChatGPT, Anthropic’s model excels in real-world coding tasks, particularly handling complex codebases, outperforming OpenAI’s o3-mini in certain programming benchmarks.

OpenAI is now following suit, planning to merge its GPT and reasoning models into a single AI. However, pricing remains a key question, as cheaper alternatives like DeepSeek and Google’s Gemini models could drive costs down. The AI market is evolving into two segments—cutting-edge reasoning models with strong pricing power and a competitive landscape of “good-enough” models racing to lower costs.

STARTUPS

OpenAI roadmap update for GPT 4.5 and 5

OpenAI CEO Sam Altman announced a shift in the company’s AI roadmap, halting the standalone launch of the o3 reasoning model to focus on a unified GPT-5 system. The upcoming GPT-4.5, set to release in the coming weeks, will be the last non-chain-of-thought model before OpenAI integrates its GPT and o-series models into a single AI system with dynamic task prioritization. GPT-5, launching in a few months, will incorporate the reasoning capabilities of o3 while streamlining product offerings to enhance user experience.

Altman emphasized OpenAI’s goal of making AI "just work" without requiring users to navigate complex model selections. The free version of ChatGPT will provide unlimited access to GPT-5 at a standard level, while Plus and Pro subscribers will access higher intelligence tiers.

BIG TECH

HPE’s ships NVIDIA Grace Blackwell system

Hewlett Packard Enterprise (HPE) has shipped its first Nvidia Blackwell-based solution, the Nvidia GB200 NVL72, a rack-scale system designed for large AI clusters with advanced liquid cooling for efficiency and performance. This launch is part of the Nvidia AI Computing by HPE portfolio, introduced in June 2024, aimed at supporting AI service providers and enterprises handling trillion-parameter models. The GB200 NVL72 integrates Nvidia CPUs, GPUs, networking, and software to optimize AI model training and inferencing.

Alongside this, HPE unveiled eight new ProLiant Compute Gen12 servers featuring AI-enhanced management, security upgrades, and upcoming Intel Xeon 6 processors for data center and edge workloads. CEO Antonio Neri emphasized HPE’s position as an AI leader, citing its supercomputing expertise, liquid cooling innovations, and networking capabilities, including the pending $14 billion Juniper Networks acquisition, as key differentiators in the competitive AI market.