Full Summary
This Saturday morning, the AI landscape is buzzing with new model launches, challenging both performance benchmarks and governance. Both blockchain.news and finance.biggo.com confirm that the open-weight model GLM-5.3 is setting new standards for cost-efficiency in coding. It achieved an 87.6% solve rate on the DeepSWE software engineering benchmark at just $3.99 per rollout, significantly undercutting Claude Fable 5, which costs $21.63. Meanwhile, a mysterious new AI model, Ox Alpha, is making waves. Crypto Briefing and The Next Web both report that Ox Alpha is outperforming leading models like Claude Fable 5 and GPT-5.6 Sol in coding tasks, yet its creator remains unknown. The Next Web highlights privacy concerns, especially for European businesses, as prompts and completions are retained by this unidentified provider, posing a risk under the new AI Act's transparency obligations. In other news, finance.biggo.com and TradingView confirm the open release of Moonshot AI’s Kimi K3, a 2.8 trillion-parameter model that matches top Western models with significantly less investment. TradingView adds that Harvey's new Tenet legal model, built on Kimi K3, shows an 82% improvement on legal benchmarks at a fraction of the cost of proprietary models. Google's Gemini 3.7 Flash is also breaking records, with OfficeChai reporting it as the fastest-growing model ever, now integrated into Google Search and the Gemini app. It boasts an output of 340 tokens per second, nearly three times that of GPT-5.6 Terra. What nobody expected: Startup Fortune reveals Nvidia's AVO system achieved a perfect score on the ARC-AGI-3 benchmark, not by making the AI model smarter, but by wrapping the Claude Opus 5 model in a better "harness" with persistent memory and a supervisor. Finally, Awaz The Voice reports that the next phase of enterprise AI will focus on managing continuous upgrades and costs. For instance, moving from Gemini 2.0 Flash to 3.5 Flash can increase input token pricing from 10 cents to $1.50 per million tokens. This means your company's AI budget is constantly evolving, and choosing the right model for specific business needs, not just the newest, is crucial to avoid unexpected costs.