Google Gemini 4 Argon Arrives as Wall Street Shifts to Personal Agents
Google unveiled Gemini 4 Argon on Wednesday with competitive benchmarks and aggressive pricing matching OpenAI's GPT-6.1 Sol. Yet momentum in AI is shifting to…
Google unveiled Gemini 4 Argon on Wednesday, delivering competitive benchmarks and aggressive pricing that matches OpenAI's newly discounted GPT-6.1 Sol model. Yet the company faces a harder challenge: the momentum in artificial intelligence is shifting decisively toward personal agents.
Argon ties OpenAI on a key cybersecurity test and posts leading results in software engineering, according to industry benchmarks. Pricing starts at $2 per million input tokens and $10 per million output tokens — the same rate OpenAI set for GPT-6.1 Sol. A token is equivalent to about three-quarters of one word.
The problem for Google is where investor and consumer attention has moved. Meta's free Muse app, launched last month, soared to the top of Apple's App Store, surpassing ChatGPT. OpenAI this week rolled out Dots, its personal agent offering, capitalizing on demand for tools that can book travel, sift emails, and organize expenses on a user's behalf. Meanwhile, Google's Spark remains paywalled, even though the company has access to billions of users and an extensive ecosystem of personal data.
Google's Gemini product chief told CNBC the company is exploring whether Argon could power more complex tasks within Spark. That exploration underscores a harder truth: building a faster model matters less than building something people actually use.
Internal doubts about coding performance
Inside Google, skepticism lingers about how well Gemini 4 performs in real-world applications. Employees with direct access to the model report that it struggles with certain coding tasks, even as it delivers solid benchmark results, according to Bloomberg. The company had planned to release a different version, Gemini 3.5 Pro, in June but abandoned the effort.
Google head of DeepMind Koray Kavukcuoglu defended the model's performance at a conference last week. "I have the utmost trust in the team," he said. "In my mind, it's a certainty that we are always gonna be at the frontier."
Opinion inside Google is split. Some employees believe Anthropic's Fable and OpenAI's Astra models are improving faster than Gemini and will outpace it in certain areas. Others contend Gemini 4 has caught up with leading labs. Google said it would be inaccurate to characterize Gemini 4 as underperforming in coding.
The personal agent race heats up
The shift away from frontier model development reflects a broader industry pivot. Anthropic CEO Dario Amodei triggered debate three weeks ago by urging the top AI labs to slow development as safety concerns intensify about potential dangers.
Over the past three months, Alphabet's stock has fallen roughly 6%, while Meta climbed 19% — a gap fueled partly by September's rally, the sharpest monthly gain since 2024. That disparity hints at where capital and consumer interest are flowing: not to the model that wins on a benchmark, but to the agent that saves time.
Google is also dismantling Gemini's Gems feature, which let users build custom AI assistants for specific tasks. Starting November 17, Gems will be automatically migrated to "skills" that work across different AI workflows. The shift reflects Google's struggle to keep users engaged with fragmented AI products as competitors consolidate around simpler, more powerful agents.
Source: cnbc.com
