AI Trends — 5 June 2026
Anthropic Tops OpenAI, Eyes $1 Trillion Valuation
In the biggest AI business story of the week, Anthropic has officially become the most valuable AI startup in Silicon Valley. The company closed a $65 billion Series H financing round at a $965 billion valuation, surpassing OpenAI. Alongside the announcement, Anthropic confirmed it has confidentially filed for an IPO.
The company's revenue trajectory is equally striking: Anthropic reported a $47 billion annual revenue run rate, up from $30 billion earlier this year and $10 billion for all of last year. Much of this growth is attributed to Claude Code, its AI coding assistant, which has become a dominant force in the developer tooling market.
Google Gemini 3.5 Flash Goes Generally Available
Google made Gemini 3.5 Flash generally available this week, positioning it as a frontier-level model at 4x the speed of comparable alternatives. Key specs:
- Pricing: $1.50 / $9 per 1M tokens (input/output)
- Context window: 1M tokens
- Performance: 76.2% on Terminal-Bench 2.1, beating Gemini 3.1 Pro on both coding and agentic tasks
The launch is part of Google's broader push into agentic AI, which also includes the unveiling of Antigravity 2.0, its new agentic framework aimed at winning over developers.
New Model Benchmarks: Claude Opus 4.8 & GPT-5.5 Instant
This week brought fresh performance data on the latest frontier models:
- Claude Opus 4.8 (updated May 28): Scores 88.6% on SWE-bench Verified and 74.6% on Terminal-Bench 2.1, with an Elo rating of 1890 on GDPval-AA. The model introduces dynamic parallel-subagent workflows and a 2.5x fast mode.
- OpenAI GPT-5.5 Instant: Continues to set competitive benchmarks, reinforcing OpenAI's position as a primary contender despite Anthropic's valuation lead.
The AI Coding Wars Heat Up
Microsoft and Google are aggressively moving to challenge Anthropic and OpenAI's dominance in AI coding tools. Both companies are leveraging their cloud infrastructure — Azure and Google Cloud respectively — to offer developers compelling alternatives to Claude Code and Copilot. Executives across all four companies describe the frontier race as effectively neck-and-neck, with each lab making different tradeoffs on cost, speed, and compute.
Training Efficiency Milestone: 100B Parameters for $1.25/Hour
In a notable efficiency breakthrough, Orion-100B demonstrated that a 100-billion-parameter model can now be trained at $1.25 per hour — a cost point that was unthinkable just 18 months ago. This development signals a fundamental shift in who can afford to build large-scale models, with implications for startups, research institutions, and enterprise AI teams alike.
The Big Picture: Agentic AI Is Now Operational
Across the industry, the conversation has shifted from "what can AI do?" to "what workflows is AI running?" Agentic systems — AI that completes multi-step tasks autonomously — are moving into production across sales, legal, research, software development, and customer support. The defining characteristic of this wave is not novelty but operational integration: AI agents embedded in real business processes, under human review, doing real work.
Vertical AI specialization is accelerating alongside this trend, with purpose-built models for oncology, legal review, finance, and CAD workflows gaining traction. Tempus, for example, upgraded its Lens clinical AI platform to use agentic workflows for oncology drug development.
Sources: CNBC · Invezz · CNBC Coding Wars · LLM Stats · AIApps · O'Reilly Radar