GPT-5.2 vs Gemini 3 Flash: The Week AI Competition Peaked

Nvidia Acquires SchedMD to Strengthen AI Infrastructure

Nvidia completed its acquisition of SchedMD, the company behind the popular Slurm workload scheduler used extensively in HPC and AI training clusters. The deal, announced December 15, 2025, sees Nvidia pledging to keep Slurm open-source while integrating it more deeply into their AI infrastructure tooling. This strategic move complements Nvidia's hardware business by giving them control over a critical piece of distributed computing software that orchestrates workloads across thousands of GPUs. The acquisition strengthens Nvidia's position as the end-to-end provider for AI infrastructure, from chips to scheduling software.

U.S. Opens National Labs to Major AI Companies

The U.S. government announced updated policies on December 19 expanding access to national laboratory computing infrastructure for companies including OpenAI, Google, and Nvidia. The policy change allows these AI leaders to leverage exascale computing resources and research datasets previously reserved for government and academic use. This move is part of a broader strategy to maintain U.S. competitiveness in AI research and development. Industry analysts see this as a significant shift that could accelerate large-scale AI training and research initiatives.

AI Market Momentum Continues Strong Into Year End

Financial markets showed continued bullish sentiment around AI investments throughout the week of December 16-20. Nvidia demand signals remained strong with Blackwell chip orders exceeding expectations, while Big Tech companies signaled increased AI infrastructure spending for 2026. Enterprise AI adoption metrics continue trending upward, with analysts projecting AI will represent an increasingly significant portion of corporate IT budgets. The market appears to be pricing in continued AI growth despite broader economic uncertainties.

OpenAI Launches GPT-5.2 with Codex Specialization

Following the general GPT-5.2 release on December 11, OpenAI launched the Codex variant on December 18 specifically optimized for large-scale code refactoring and cybersecurity applications. The new model shows significant improvements in understanding complex codebases, identifying security vulnerabilities, and suggesting comprehensive fixes. GPT-5.2 Codex is designed for enterprise development teams working on legacy system modernization and security audits. Early benchmarks suggest 40% improvement on code understanding tasks compared to GPT-5.1.

Google Releases Gemini 3 Flash for Efficient Agentic AI

Google launched Gemini 3 Flash on December 17, positioning it as the most cost-efficient frontier model for agentic workflows. The model delivers reasoning capabilities comparable to larger models while significantly reducing inference costs and latency. Google made Gemini 3 Flash the default model in the Gemini app, signaling confidence in its capabilities for everyday users. The release includes native support for function calling and tool use, making it particularly suited for AI agent applications that need to interact with external systems and APIs.

DOE Genesis Mission AI Partnerships Gain Traction

The Department of Energy's Genesis Mission, launched in November, saw significant progress this week with new collaboration agreements signed around December 18. The initiative brings together national laboratories, AI companies, and research institutions to advance scientific computing with AI. Projects span areas from climate modeling to materials science, leveraging the combination of exascale computing resources and modern AI techniques. The partnerships represent a new model for public-private collaboration in AI research.

OpenAI

Anthropic

Google