Anthropic posts first profit; Meta opens 30 B agent model
Anthropic reports its first operating profit, Meta releases a 30-billion-parameter open model, and AI tools design novel viruses published in Science.
By BINA Editorial
Today's brief spans economics, open-source releases, model rankings, biosecurity, and geopolitics — six developments that together mark a shift in how AI is built, financed, and governed.
Anthropic reports its first-ever operating profit
Anthropic posted a Q2 2026 operating profit of $559 million on $10.9 billion in revenue — making it the first frontier AI laboratory to reach operating profitability, two years ahead of its own internal schedule. Revenue more than doubled from Q1's $4.8 billion, driven by enterprise adoption of Claude and strong developer API demand. The milestone matters beyond the balance sheet: it demonstrates that a safety-focused laboratory can sustain itself without indefinite subsidy, a case Anthropic's leadership has long needed to make.
Meta releases Muse Glimmer, a 30-billion-parameter open-weight model
Meta published Muse Glimmer, a 30-billion-parameter multimodal agent model released under the Apache 2.0 licence, meaning anyone can use, modify, and redistribute it without restriction. The model is designed to run on a single consumer-grade GPU, targeting coding assistance, tool use, and always-on local agents that operate entirely without a cloud connection. The release continues Meta's strategy of publishing competitive open-weight models, putting meaningful AI capability directly into the hands of individuals, researchers, and small organisations that cannot afford proprietary API costs.
Claude Opus 5 leads the August 2026 model rankings
The Artificial Analysis Intelligence Index for August 2026 places Claude Opus 5 at the top with a composite score of 63, edging out Anthropic's own Fable 5 (62) and OpenAI's GPT-5.6 Sol (61). Google's Gemini 3.1 Pro trails the leaders at 48 after several months without a new flagship model release. The rankings confirm an increasingly compressed field at the frontier, where just a few points on a composite benchmark — spanning reasoning, coding, and instruction-following — separate the leading models.
Gartner warns: cheaper tokens will not lower the AI bill
A new Gartner forecast projects that the cost per agentic AI workflow will rise more than fivefold between now and 2028, even as the price per individual token falls by roughly 95 percent by 2030. The explanation lies in how agentic systems are used: longer context windows, more capable models, and multi-step reasoning chains mean that each workflow consumes far more compute than a simple completion call. Gartner's analysis is a direct caution for product and finance teams who are budgeting AI expenditure on the assumption that falling token prices will translate into falling total bills.
AI-designed bacteriophages published in Science; biosecurity debate follows
Researchers at Stanford University and the Arc Institute used AI genomic models to design novel bacteriophages — viruses that target bacteria — entirely from scratch, with results published in the journal Science. The engineered phages outperformed natural relatives at overcoming bacterial resistance, a result with potential significance for treating antibiotic-resistant infections — one of medicine's most pressing problems. Biosecurity experts cited in the coverage warn that the same generative techniques could in principle be adapted to design harmful biological agents, and they are calling for governance frameworks that keep pace with the underlying capability.
Trump-backed crypto firm's AI venture includes models restricted by US policy
World Liberty Financial, the cryptocurrency firm associated with the Trump family, has partnered with a platform that offers AI models from Chinese companies the US government has separately flagged as national-security concerns — including Alibaba, Baidu, and Z.ai. Forty-three of the ninety models available through the partner platform originate from those firms. The arrangement illustrates a live tension between commercial AI partnerships and export-control policy, and is expected to draw scrutiny from congressional and regulatory bodies.