insidejob

March 2026: the densest model release window in AI history

March and early April 2026 produced one of the densest model release windows in AI history. Three frontier models dropped in a single month:

  • GPT-5.4 (March 5) — Standard, Thinking, and Pro variants. Record 83% on GDPval, record scores on OSWorld-Verified and WebArena Verified computer-use benchmarks.
  • Gemini 3.1 Ultra — Native multimodal reasoning. Gemini 3.1 Pro leads SWE-bench Verified at 78.80% and scores 94.3% on GPQA Diamond.
  • Grok 4.20 Beta 2 — Enhanced real-time web access from xAI.

Meanwhile, the open-source side kept pace with DeepSeek V4 (1T parameters on OpenRouter), Qwen 3.6 Plus (1M native context), and Google’s Gemma 4 (26B params, runs on consumer hardware).

LLM Stats tracked 255 model releases from major organizations in Q1 2026 alone.