Meta, OpenAI, Google: Who Will Win the AGI Race?

At ASSIST Software, innovation happens at the intersection of leading research and real-world application. Nowhere is that more visible than in the global race toward Artificial General Intelligence (AGI),  the dream of creating AI that can think, learn, and adapt like humans.

This isn't science fiction anymore. Tech giants like Meta, OpenAI, Google, Anthropic, and a fast-growing roster of Chinese AI labs are pouring billions into infrastructure, talent, and research. The race is accelerating, and the stakes are massive, shaping economies, global security, and how technology integrates into our daily lives.

Meta's AGI ambitions are no longer just ambitious talk. Since establishing Meta Superintelligence Labs in mid-2025, the company has been through an internal reset. Alexandr Wang, former CEO of Scale AI, joined as Chief AI Officer to lead the new unit, a $14.8 billion deal that gave Meta a 49% stake in Scale AI in the process.

The once-speculative Behemoth model has had a turbulent journey: Meta completed training but delayed its release due to underwhelming internal performance. Wang's team subsequently scrapped the model and started fresh and, in a striking philosophical shift, began exploring whether the next flagship model should be closed-source, breaking with Meta's long-standing open-source strategy through the Llama family. Mark Zuckerberg has not officially signed off on that change, and the company maintains that its open-source commitment is "unchanged", but the internal debate signals how much the calculus has shifted as Meta races to monetize its AI investments.

Meanwhile, Meta's newest publicly available model, Muse Spark, is already live across the Meta AI app, WhatsApp, Instagram, Facebook, and Messenger, featuring multimodal reasoning capabilities and strong performance in science, math, and health queries.

The compute ambitions remain staggering. Clusters codenamed Prometheus and Hyperion, the latter projected at up to five gigawatts, continue to be built out, backed by capital expenditure commitments of up to $72 billion in 2025 alone.

For ASSIST Software, we watch developments like these closely, not just as industry news, but as signals for how advanced AI capabilities will filter down into tools we can adapt and integrate for our clients.

When OpenAI unveiled GPT-5 in August 2025, it was positioned as the company's most capable model yet: stronger reasoning, fewer hallucinations, better performance in specialized domains. Nine months on, the story has moved fast.

By April 2026, OpenAI had shipped GPT-5.5, described by the company as its "smartest and most intuitive" model yet, arriving just six weeks after GPT-5.4. The release cadence underscores how fiercely frontier labs are now competing, model versioning has shifted from landmark events to a near-continuous cycle.

GPT-5.5 excels particularly in agentic coding, knowledge work, and long-horizon task execution, designed to handle complex, multi-part tasks with minimal hand-holding. It also ships with OpenAI's strongest safety profile to date. A dedicated variant, GPT-5.5-Cyber, is rolling out in limited preview to vetted cybersecurity teams.

In May 2026, OpenAI released GPT-5.5 Instant as the new default model for all ChatGPT users, replacing GPT-5.3 Instant. The platform now reports over 900 million weekly active users and 50 million subscribers, though OpenAI acknowledges it is also managing a growing narrative that rival Anthropic has been making headway in the enterprise market.

As for GPT-4o, it was formally deprecated in February 2026, despite user petitions.

Google’s DeepMind's Gemini family has moved significantly since mid-2025. The company shipped Gemini 3, now at version 3.1 Pro, positioning it as state-of-the-art on reasoning, multimodal understanding, and agentic coding. Gemini 3 Deep Think achieved an unprecedented 84.6% on the ARC-AGI-2 benchmark and gold-medal-level performance on the 2025 International Math Olympiad.

Strategically, Google is pivoting Gemini from a chatbot to an operating layer. At the Android Show in May 2026, the company unveiled Gemini Intelligence for Android, integrating Gemini into Chrome, Autofill, app automation, and Android Auto across Samsung and Pixel devices, before a wider rollout later in 2026. The company's vision, as Android chief Sameer Samat put it: "We're transitioning from an operating system to an intelligence system."

DeepMind CEO Demis Hassabis continues to articulate an ambitious AGI roadmap, focused on "root node" problems in fusion energy, material science, and world models. The company also released Deep Research Max in April 2026, built on Gemini 3.1 Pro, offering enterprise-grade autonomous research agents for finance, life sciences, and market research.

While Meta, OpenAI, and Google dominate the AGI headlines, Anthropic has quietly become one of the most consequential players in the race, on its own terms. The company's founding premise is unusual: it openly acknowledges it may be building one of the most dangerous technologies in history, and presses on anyway, on the logic that if AGI is coming regardless, it is better to have safety-focused labs at the frontier than to cede that ground to others.

Commercially, the results have been striking. Anthropic's annualized revenue climbed from $9 billion at the end of 2025 to $30 billion by April 2026, growth Amodei himself described as "just crazy" and "too hard to handle," having planned for 10x growth and received 80x instead.

On AGI timelines, CEO Dario Amodei has been the most direct of any major lab leader. At Davos in January 2026, he predicted AGI-level AI within two years and warned that up to 50% of white-collar jobs could be disrupted within one to five years. Where others in the race talk about winning it, Amodei talks about managing it.

For businesses, Anthropic's rise is a useful reminder that the AGI race is not just a contest between the biggest spenders. Differentiated positioning, in this case, around trust and safety, is proving to be a genuine competitive advantage at the frontier.

The race extends well beyond Meta, OpenAI, and Google, and the biggest story is DeepSeek.

In January 2025, the Hangzhou-based Chinese startup released its R1 reasoning model, which matched or outperformed leading U.S. models on major benchmarks, reportedly built in just two months for under $6 million using lower-capacity chips. The release rattled global tech markets and raised serious questions about the U.S. compute-spending thesis. Time Magazine named it one of the "Best Inventions of 2025."

In April 2026, DeepSeek released a preview of V4, its most significant update since R1, featuring major upgrades in reasoning and agentic capabilities. Unlike R1, V4's market impact was more muted; investors have already priced in the reality that Chinese AI is competitive and cheap. But the competitive pressure it represents on U.S. labs is now a structural feature of the landscape, not a surprise.

Meanwhile, the White House's science and technology office has accused foreign entities, primarily Chinese ones, of large-scale efforts to extract knowledge from U.S. frontier AI models.

The AGI race is shaped as much by personalities as by corporate strategy. The rivalry between Sam Altman and Elon Musk continues, now with the added texture of Musk's own xAI competing directly with Grok models, while Altman has also had to contend with a wave of leadership departures to Meta's Superintelligence Labs.

Alexandr Wang's move from Scale AI to Meta was the defining talent event of the past year. And the broader pattern, prominent researchers leaving big labs for leaner startups or rival giants, has only accelerated, reinforcing that AGI's future may be shaped as much by individual bets as by trillion-dollar campuses.

The fundamental uncertainties haven't changed. Experts still debate what AGI actually means, some require matching human intelligence across all domains, others demand autonomous learning in entirely new areas. Most frontier labs now speak more confidently about timelines than they did a year ago, but claims vary wildly.

Risks remain: bias, misinformation, misuse in cybersecurity and geopolitics, and the economic disruption of AI taking on more complex roles. The rapid release cadence of models, sometimes just weeks apart, also raises questions about whether safety evaluation is keeping pace. The agentic turn, where AI models take real-world actions on behalf of users, introduces an entirely new risk surface that regulators are only beginning to engage with.

The AGI race shapes the headlines, but the more immediate question for most businesses is simpler: where does my industry stand right now? The honest answer is that adoption is wide but shallow. The gap between experimenting with AI and transforming a business with it is wide.

According to a survey by Deloitte, worker access to AI rose 50% in 2025, and the number of companies with significant AI deployments in production is set to double within six months, yet just 34% of leaders say they are truly reimagining their business around AI, rather than simply applying it to existing workflows.

The next frontier is the shift from tools to agents.

For businesses, the AGI race is a glimpse into the near future of work and competition. The pace has only accelerated since 2025. Tools that were once confined to research labs are shipping in consumer products. Preparing now means:

• Experimenting early with AI to automate workflows and enhance decision-making
• Upskilling teams to work alongside advanced AI agents, not just AI chatbots
• Building adaptable tech stacks: model generations are now rotating every few months, not every few years
• Leading with trust by setting ethical and transparency standards in advance, especially as agentic AI begins to take autonomous action

Organizations that treat AGI readiness as a strategic priority will have a meaningful competitive edge.

For ASSIST Software, the AGI race is an opportunity. Big tech may be building the models, but agile innovators like us are the ones who will turn them into real-world solutions that solve client problems today.

We have integrated advanced AI into automation, analytics, and customer experiences, and hold our official AI certification recognizing our expertise and commitment to trusted, high-quality AI solutions. We invest in continuous learning, research partnerships, and hands-on experimentation, and we follow the frontier closely enough to adapt as the models beneath us keep moving.

We lead with ethics and transparency, because in an era where AI is taking on more autonomous roles, trust will be as critical as capability.

We’re ready to be the bridge for our clients and partners, guiding them through AI adoption now and preparing them for the transformative possibilities that AGI will bring tomorrow.

The race toward AGI is on. At ASSIST Software, we’re preparing to help you win.

Follow our page for more insights on AI, AGI, and the future of technology.