Summary

This article summarizes significant developments in large language models (LLMs) over the past six months, highlighting key advancements and model transitions. It reflects on the November 2025 inflection point that marked a turning point for coding agents and their capabilities.

Key Points

• The "best" LLM model changed hands five times among major providers in November 2025.
• Claude Sonnet 4.5 was initially the leading model, released on September 29, 2025.
• The sequence of leading models included GPT-5.1, Gemini 3, GPT-5.1 Codex Max, and Claude Opus 4.5.
• Coding agents improved significantly, transitioning from "often-work" to "mostly-work" due to advancements in Reinforcement Learning from Verifiable Rewards.
• The OpenClaw project emerged as a notable personal AI assistant in February 2026.
• Google released the Gemma 4 series, while GLM introduced the GLM-5.1 model, a 1.5TB open weight model.
• Qwen released Qwen3.6-35B-A3B, a laptop-compatible model that exceeded expectations in performance.

Analysis

The developments in LLMs over the past six months indicate a rapid evolution in AI capabilities, particularly in coding applications. The shift towards more reliable coding agents suggests that organizations can increasingly rely on AI for real-world programming tasks, enhancing productivity.

Conclusion

IT professionals should explore the latest LLMs and coding agents to leverage their capabilities in software development. Staying updated on these advancements can lead to improved efficiency and innovative project implementations.