I had to test it more:

solve: 5.9=x+5.11

gemini : FAIL (-0.2)
perplexity : FAIL (-0.21)
ChatGpt5 : FAIL (-0.21)
Grok3 : PASS (0.79)
ASI:ONE : PASS (0.79)
Claude Sonnet 4's answer was so ridiculous I had to screencap it