Qwen4-Max aces coding, long-document and finance tests; long context is its edge
Div_pradeep · x · 2026-11-18
Divpradeep's verdict: Qwen4-Max did well on all three tests (coding, long documents, finance), with long-context work standing out — feeding a 200-page report and getting page-referenced analysis back is genuinely useful. He suggests judging it yourself with a long document you already know well.
More from Models
- AI agents sometimes produce unreadable but correct math — burny_tech · 2026-11-18
- Code model eval: Rust hits 95-100% accuracy as 13 languages see big gains — shuding · 2026-11-18
- Asked Opus to make a video explaining consciousness and existence — "no wiser but cool" — flowersslop · 2026-11-18
- theo retests GPT-6.2 Astra: better, but still 'directionally very bad' at debugging — airesearch12 · 2026-11-18
- Claude correctly folds ubiquitin from a plain prompt, no specialized tools needed — iScienceLuvr · 2026-11-18
- GPT-6.2 Astra Tops Bug-Hunt Coding Benchmark at 62.7/105, One-Fifth Sonnet's Cost — PawelHuryn · 2026-11-18