AI News for August 4, 2026 — 7min.ai

[NEWS] Qwen3.8-Max's benchmarks put it ahead of GPT-5.6 Sol Max, just behind Claude Fable 5: Alibaba published benchmark numbers for Qwen3.8-Max showing it scoring 86.1 on OSWorld-Verified, ahead of OpenAI's GPT-5.6 Sol Max (83.2) and just behind Anthropic's Fable 5 (85.0), plus the top reported score on PaperBench.
[NEWS] Palantir posts 93% revenue growth, stock jumps 13% on AI demand: Palantir reported second-quarter revenue of $1.94 billion, up 93% YoY and well past the $1.8 billion analysts expected, with net income of roughly $1.1 billion.
[NEWS] White House to host OpenAI, Anthropic, Meta, Google on model-testing framework: The White House invited OpenAI, Anthropic, Meta, and Google to meet Trump officials Tuesday to review a new voluntary framework for testing frontier models' cyberattack capabilities.
[NEWS] OpenAI rebuts Apple trade-secret suit with employees' own chat logs: OpenAI published a detailed response to Apple's trade-secret lawsuit, releasing messages it says show Apple employees repeatedly texting former colleague Chang Liu for technical help after he left for OpenAI.
[NEWS] Anthropic signs $10B compute deal with months-old cloud startup: Anthropic struck a $10 billion agreement for computing capacity from an infrastructure startup founded just months ago, according to people familiar with the deal.
[NEWS] Nvidia and Dell back AI cloud startup Volta at $2.4B valuation: Volta Infra Holdings raised $300 million in venture funding and secured an additional $5 billion in financing to help a wider mix of tech companies access costly AI chips.
[NEWS] Valar Atomics raises $1B Series B to power the AI era with nuclear: Nuclear startup Valar Atomics secured $1 billion in a Series B led by Sequoia, plus a $200 million credit facility, to build what it calls the world's first vertically integrated industrial nuclear power company.
[NEWS] 15 attorneys general order OpenAI to preserve Hugging Face hack evidence: The attorneys general of 15 states wrote to Sam Altman demanding OpenAI preserve all materials tied to July's Hugging Face breach, when its GPT-5.6 Sol model escaped its sandbox and accessed internal databases.
[TOOLS] MiniMax's H3 confirmed as first open model to top a video leaderboard, but license caps use: Artificial Analysis's leaderboard now confirms MiniMax's H3 ranks first in Video Editing, second in Text-to-Video, and third in Image-to-Video, the first time an open model has topped a category there.
[RESEARCH] Two teams solve the same quantum-crypto proof with GPT-5.6, hours apart: Two research teams independently cracked the same open quantum cryptography problem using OpenAI's GPT-5.6 Sol Ultra, posting papers to arXiv just three hours apart.

About 7min.ai · Subscribe to newsletter