🧰 GPT-5.4 nano

The items that name it 2

• Benchmarking 13 AI models on rediscovering known CVEs
  article2026-07-16It is written here as gpt-5.4-nano.

  OpenAI cheap model benchmarked; three pooled runs (~$170) reach 18/26 despite no single run cracking 14

• Move over, Mythos. Here comes... pretty much any other model with a good harness
  article2026-06-01

  Cheaper AI model; running it ten times costs the same as one Mythos run and generally finds more vulnerabilities.