01
GLM-5.3 Ships as Open-Weight Coding and Cybersecurity Model from Z.ai
Z.ai released GLM-5.3 on August 14, built on the same base as GLM-5.2 with extended post-training. It scored 84.5 on the CyberGym cybersecurity benchmark, edging out Anthropic's Mythos 5 by 0.7 points, and enters the best-available ranking tied for 11th of 681 models. API pricing is $1.40 per million input tokens and $4.40 per million output tokens, significantly cheaper than Grok 4.6 at $8 combined. On Terminal-Bench 3.0 and DeepSWE, it still trails Claude Fable 5 and GPT-5.6 Sol.
02
Mistral Begins Hosting Z.ai's GLM-5.2 in Push for European AI Sovereignty
Mistral is now hosting third-party open models for the first time, starting with GLM-5.2 from Z.ai (Zhipu), a Chinese lab on the US entity list. GLM-5.2 offers a 1-million-token context window at $1.40 per million input tokens with open weights and no usage restrictions. ASML, Amadeus, Capgemini, Caisse des Depots, and CMA CGM signed multi-year compute commitments for Mistral's planned 1GW European data center buildout by 2030.
03
Ornith-1.5: Open-Source Coding Models Claim Near-Claude Opus 4.8 Performance
Ornith released three new open-source coding models, including a 397B parameter mixture-of-experts model and a 9B parameter model with quantized builds for local deployment. The lab claims its largest system approaches Claude Opus 4.8 on coding benchmarks, though the comparison is vendor-reported. The 9B model is the practical release for developers, runnable on local hardware with quantized builds.
04
Alibaba Open-Sources Qwen3.8 Max Weights as Cloud Revenue Hits 45% Growth
Alibaba Cloud reported 45% external revenue growth in Q1 2027, a 22-quarter high, with adjusted EBITDA margin of 11.6%. The company open-sourced the weights of its flagship Qwen3.8 Max model after three months, and also launched Qwen3.8-27B, a 27B parameter multimodal model under Apache 2.0 license. Alibaba Cloud cut hyperscale AI data center delivery time to 100 days.
05
Broadcom Seeks $60B+ in Debt Financing for AI Chip Procurement
Broadcom is in talks with multiple lenders to raise more than $60 billion in debt to support chip procurement and computing infrastructure, adding to a rush of deals aimed at financing AI infrastructure. The financing comes as AI chips increasingly become an investable asset class, with billionaire investor Jeff Gundlach warning it could signal a market peak.
06
Data Center Capex Forecast to Surpass $3 Trillion by 2030
Dell'Oro Group forecasts worldwide data center capital expenditure will surpass $3 trillion by 2030, with high-end accelerators powering AI-optimized servers representing the largest share. General-purpose server demand is expected to benefit from growing inference, agentic AI, and storage workloads, broadening infrastructure growth beyond accelerated computing.
07
Vals AI Raises $40M from a16z for Independent AI Benchmarking
Vals AI closed a $40 million Series A at a $400 million valuation, led by Andreessen Horowitz, to expand independent AI benchmarks across coding, cybersecurity, and professional domains. The company's benchmarks revealed that frontier models fail 52% of real finance analyst tasks, highlighting the gap between vendor-reported performance and real-world capability.