OpenAI released GPT-5.5 Instant with 52.5% fewer hallucinations and three new realtime audio models. Anthropic partnered with SpaceX to access 220,000 GPUs. Microsoft reports 17.8% of the global workforce now uses generative AI, but the gap between rich and poor countries is widening: 27.5% adoption in the Global North versus 15.4% in the Global South.
Chapters
OpenAI's GPT-5.5 Instant goes default, Anthropic locks in 220,000 GPUs from SpaceX AI's Colossus 1, OpenAI and Anthropic launch enterprise JVs with Wall Street, and Microsoft's data shows the world is splitting into who's adopting AI and who isn't. Welcome to The Median Brief, based on The Median by DataCamp. OpenAI's new default model, GPT-5.5 Instant, hallucinates 52.5% less on complex topics. It also remembers more of your past conversations to tailor answers. On the developer side, 3 new audio models change what's possible in the API: live voice reasoning with GPT Realtime 2, live translation across 70+ languages with Realtime Translate, and low-latency transcription with Realtime Whisper.
Anthropic signed a deal this week with SpaceX AI to use the full Colossus-1 supercomputer, adding more than 300 megawatts of capacity and over 220,000 NVIDIA GPUs across H100, H200, and GB200 accelerators. The new compute flows directly to Claude Pro and Claude Max users. Anthropic doubled the 5-hour rate limits on Claude code, and significantly raised API rate limits for Claude Opus.
OpenAI and Anthropic both spun up enterprise joint ventures with Wall Street this week. OpenAI raised over $4 billion for a venture valued at $10 billion, backed by TPG and Brookfield. Anthropic partnered with Blackstone, Hellman Friedman, and Goldman Sachs on a $1.5 billion company focused on mid-sized businesses. Both will field forward-deployed engineers to embed with client teams and build custom tools.
Microsoft's Q1 2026 Global AI Diffusion Report puts generative AI usage at 17.8% of the world's working-age population, up from 16.3% in late 2025. Anonymized telemetry, adjusted for internet, population, and device share, The strongest cross-country read we have. The top of the table is locked in. UAE leads at 70.1% adoption, Singapore second at 63.4%, Norway third at 48.6%. The top 11 kept their exact positions from late 2025. Below the top tier, rankings churn. The US climbed from 24th to 21st on 31.3% adoption. South Korea jumped 18th to 16th, up 6.4 points. Taiwan moved 23rd to 20th. Denmark, Germany, and Poland gained users but still slid down. The bigger split is geographic. Generative AI reached 27.5% of the Global North versus 15.4% of the Global South. The gap is widening. North adoption is growing more than twice as fast. The bottleneck? Connectivity, Electricity, digital skills. The fastest-growing markets are almost all in Asia. 12 of the 15 top growth economies since June 2025 sit there, each adding at least 25% more users. South Korea up 43%, Thailand 36%, Japan 34%, plus Mongolia, Iran, Laos, Turkey, Vietnam, Cambodia. The driver behind the Asia wave is language. Recent models reach near-parity on multilingual benchmarks like MMLU. AI works in Korean, Thai, or Vietnamese as well as in English.
And this was the Median Brief, based on The Median by DataCamp. For the full breakdown and the deeper analysis behind this week's stories, subscribe to the newsletter using the link in the description. Thank you for watching, and I'll see you next week.
