Latest in Tech, 3 July 2026

Anthropic Ships Claude Sonnet 5 at Discount Pricing

Anthropic released Claude Sonnet 5 on June 30, positioning it as its most agentic mid-tier model to date. The company says it handles reasoning, tool use, and coding close to the flagship Opus 4.8 while carrying a 1-million-token context window and 128,000-token maximum output. Introductory API pricing runs $2 per million input tokens and $10 per million output tokens through August 31, after which it rises to $3 and $15, still well under Opus 4.8's $5 and $25. Sonnet 5 is now the default model on Free and Pro plans and is available across Max, Team, and Enterprise tiers.

US Lifts Export Controls on Fable 5 and Mythos 5

The US Department of Commerce lifted the export controls it placed on Claude Fable 5 and its more tightly restricted sibling Mythos 5, roughly two and a half weeks after imposing them on June 12. The original order followed a jailbreak that Amazon researchers found in Fable 5, a prompt that could get the model to identify software vulnerabilities and write exploit code. Anthropic trained a new classifier targeting that specific technique, which it says now blocks it in more than 99% of attempts, and began restoring Fable 5 globally on July 1 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork.

Google Pushes Gemini 3.5 Pro to July

Google confirmed that Gemini 3.5 Pro's general availability has slipped to July, after a June target came and went without a public launch. The model was announced at I/O on May 19 and has been sitting in a limited Vertex AI enterprise preview for a handful of customers. Google attributes the delay to refinements around coding, token efficiency, and long-task performance based on early tester feedback. The model is expected to ship with a 2-million-token context window and a Deep Think reasoning mode.

Meta Plans a Cloud Business to Resell AI Compute

Meta is building a cloud infrastructure business to sell access to its AI computing power and models, according to a Bloomberg report on July 1. The plan would let the company generate revenue from excess capacity sold to outside customers, putting it in more direct competition with AWS, Google Cloud, and Azure. The move reflects how much compute Meta has built out for its own AI work and its interest in monetizing the surplus rather than leaving it idle.

Next.js 16.3 Adds Persistent Build Caching

Vercel shipped Next.js 16.3 on June 29 with a set of Turbopack improvements aimed at build and development performance. The headline change brings the persistent on-disk cache, previously used to speed up next dev, to next build, so successive production builds can reuse previously computed work. The release also adds memory eviction to reduce RAM usage during long dev sessions and includes support for a Rust-based React Compiler. Vercel says the persisted build cache went through months of hardening on its own sites before release.