
Qwen3.8 Max Catches Claude Opus 4.8; Kimi K3 Leads Cheaper
Alibaba's Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index, a 10-point jump, matching Claude Opus 4.8 while Kimi K3 still scores higher for 25 percent less.
Browse all specialized articles and insights published under the AI News department.

Alibaba's Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index, a 10-point jump, matching Claude Opus 4.8 while Kimi K3 still scores higher for 25 percent less.

Amazon, Microsoft, OpenAI, Cursor, and Vercel launch Agent Plugins, an open standard defining one package format for AI agent extensions with plugin.json and MCP support.
Bloomberg reports OpenAI's device with Jony Ive is a battery-powered smart speaker without a display, hockey puck-sized, launching in 2027 at $300–400.
OpenAI updates GPT-5.6 Sol with more focused responses and a reasoning slider, and gives free users unlimited text chats via the smaller GPT-5.6 Luna.
Anthropic is building a custom silicon team to design chips that run Claude, reducing dependence on Nvidia as it races OpenAI to scale.
Microsoft's SkillOpt turns AI agent skills into a single text file (best_skill.md) transferable across models and harnesses — a Codex-trained skill lifted Claude Code from 22.1 to 81.8.
Indeed data: AI now appears in 9.4% of UK job postings (up from ~2% in 2023) while knowledge-work postings crater — a 'two-speed labor market.'
Mistral's 3B Shieldstral model checks AI inputs and outputs via natural-language yes/no questions instead of fixed categories, matching safety models 7x its size in some benchmarks — and runs locally.

FLUX 3 Video from Black Forest Labs generates Full HD clips up to 20 seconds with native audio and lip-synced dialogue in 14+ languages, with Elo rankings ahead of Gemini Omni Flash and Seedance 2.0.
Shopify reports AI-driven traffic and orders to its stores tripled year over year in Q2 — AI search works as an extra sales channel, not a Google replacement.