Qwen3.5-4B tested across 10 coding agents: 7 perfect scores, Codex CLI fails all

A Reddit user ran Qwen3.5-4B locally across 10 coding agents with identical tasks, finding 7 achieved perfect scores while Codex CLI failed all, and documented typical failure modes of small models running agents.

2026-10-21 ~ 2026-10-22 · 2 related posts

• I benchmarked 10 AI agents on a 4B local LLM: seven aced it, Codex CLI went 0/5 — GapNew4766 · 2026-10-21
• What breaks agents on a 4B model: 50 runs across 10 harnesses on Qwen3.5-4B — GapNew4766 · 2026-10-22