Daily AI RoundupsWednesday, 17 June 2026

17 June leaned on measurement and hardware. Z.ai's 744B-parameter GLM-5.2 topped the Artificial Analysis open-weights index at 51, while NVIDIA said Blackwell swept MLPerf Training v6.0, fastest at scale and per accelerator, and expanded its AI Factory with HPE and the Vera CPU. On evaluation, OpenAI discussed better evals as benchmarks saturate or get gamed, Anthropic shared an economic framework for tracking Claude Code usage and value, and Interconnects reviewed Olmo-style post-training recipes. Snowflake shipped Adaptive Compute GA on AWS, Wolfram launched Version 15 with AI assistants, GitHub began retiring GitHub Models for new orgs, and Midjourney teased its first hardware project.

• Benchmarks
• AI chips
• Model releases
• Evaluation
• Dev tooling

Stories in this roundup

1. 01
   Z.ai's GLM-5.2 tops Artificial Analysis open-weights index at 51 with 744B parametersARTIFICIAL ANALYSIS
2. 02
   GitHub cuts off GitHub Models for new orgs on June 16, keeping existing users until retirementGITHUB
3. 03
   NVIDIA says Blackwell swept MLPerf Training v6.0, ranking fastest at scale and per acceleratorNVIDIA
4. 04
   HPE, NVIDIA expand AI Factory with Vera CPU, Agent Toolkit for production agentic AINVIDIA
5. 05
   Snowflake ships Adaptive Compute GA on AWS for spiky AI and analytics queriesSNOWFLAKE
6. 06
   Wolfram launches Version 15 with AI assistants for Wolfram Language and Mathematica notebooksSTEPHEN WOLFRAM
7. 07
   Interconnects reviews Olmo-style frontier post-training recipes with Finbarr TimbersINTERCONNECTS
8. 08
   OpenAI discusses better evals for measuring and forecasting model progress as benchmarks saturate or get gamedOPENAI
9. 09
   Anthropic shares economic research framework for tracking Claude Code usage, task value, and scaling patternsANTHROPIC
10. 10
    Midjourney will announce its first hardware project at a 6 p.m. PT San Francisco livestreamMIDJOURNEY