Wednesday, June 10, 2026

Claude’s New Model Split

Claude’s New Model Split

Today’s Overview

Good morning, Anthropic is splitting its newest Claude into a broadly available model and a more restricted cyber-focused version, while Apple is taking another swing at Siri with deeper systemwide AI. Perplexity also has fresh data on how agents shift people from search-and-click work into higher-level supervision. Let’s dive in.

Top Stories

Apple Rebuilds Siri Around Systemwide AI

Apple introduced Siri AI, a rebuilt assistant layer across iOS 27, macOS Golden Gate, and visionOS. The update connects Siri more deeply to Spotlight indexing, App Intents, View Annotations, Shortcuts, Photos, Passwords, and Apple’s Private Cloud Compute infrastructure.

  • Siri AI can draw on personal context across messages, emails, photos, and third-party apps that integrate with Spotlight.
  • The assistant is being added to Spotlight and context menus on iPad and Mac, with Vision Pro support through a spatial 3D Siri interface.
  • Developer testing begins across iOS 27 and related platforms with user beta availability planned later this year for supported devices and languages.

Anthropic Launches Claude Fable 5 and Mythos 5

Anthropic announced Claude Fable 5 for general availability and Claude Mythos 5 for selected cyberdefenders and infrastructure providers. The models are positioned as major upgrades across software engineering, research, vision, and cybersecurity, with stricter safeguards around Fable 5.

  • Fable 5 is described as strongest on longer, complex tasks where Anthropic says its lead over earlier Claude models grows most clearly.
  • Mythos 5 uses the same underlying model but has some safeguards lifted for approved cyberdefenders through Project Glasswing and related trusted access programs.
  • Both models are priced at $10 input and $50 output per million tokens, less than half the listed price of Claude Mythos Preview.

Moonshot Targets a $30B Valuation

Moonshot is reportedly preparing to raise between $1 billion and $2 billion at a valuation of about $30 billion. The financing would reflect growing investor demand around the company’s Kimi models and China’s broader AI startup race.

  • The target would be about 50 percent higher than Moonshot’s reported $20 billion valuation from its previous financing round in May.
  • If completed, the round would place Moonshot among China’s most valuable AI startups as competition among domestic model developers intensifies.
  • The company behind Kimi has seen its valuation rise by more than sixfold from the level reported at the end of last year.

Research & Analysis

Perplexity Maps How Agents Reshape Work

Perplexity and Harvard Business School compared its Computer platform with Search to study how agents change knowledge work. The study found that users ask agents for more complex, creative, and execution-heavy tasks than they ask of search tools.

  • Across matched sessions, Computer performed 26 minutes of machine execution on average, compared with 33 seconds for Search.
  • Estimated task time fell from 269 minutes to 36 minutes when comparing a Search plus human workflow with a Computer plus human workflow.
  • Computer queries required higher-order cognition 76 percent of the time, compared with 55 percent for Search queries.

Trending AI Tools

  • North Mini Code A 30B-parameter MoE coding model with 3B active parameters, released under Apache 2.0 for agentic software development in sovereign AI settings.

  • Hero Studio Photos Turns one product photo into listing-ready shots from multiple angles.

  • Archyl Generates interactive C4 architecture diagrams and drift metrics directly from repository code.

Quick Hits

  • Xiaomi MiMo UltraSpeed hits 1,000 tokens per second on a standard 8-GPU commodity node, with a limited API trial running from June 9 to June 23.

  • New York AI ad disclosures make the state the first in the U.S. to require disclosure of AI-generated actors in ads, with $1,000 fines.

  • ChatGPT interactive charts add inline graphs from data directly inside the chat flow.

  • Kimi Work runs up to 300 parallel local desktop agents for browser tasks and financial market workflows.

Keep reading for free

Enter your email. If you're already subscribed, we'll send a sign-in code. If not, you'll subscribe in the next step.

Free access. Subscribe once, then use the same email on future issues.

Free to read. Subscription just unlocks the full issue.