Frontier AI Models February 2026: GPT-5.3 Codex, Claude 4.6, Gemini 3.1, GLM-5, DeepSeek

Summary

February 2026 saw multiple frontier model releases. OpenAI shipped GPT-5.3 Codex (Feb 5) with 77.3% on Terminal-Bench and improved cybersecurity preparedness. Anthropic released Claude Opus 4.6 with Agent Teams, 80.8% on SWE-bench Verified, and 1M-token context in beta. Zhipu GLM-5 (744B params, Huawei Ascend–trained) reached top open-weight on Artificial Analysis at $0.11/M tokens. Google's Gemini 3.1 Pro reached 77.1% on ARC-AGI-2. DeepSeek V3.2 expanded to 1M+ context; Moonshot Kimi K2.5 (1T