Sonnet 5.5 effort settings make no difference in 15-task coding test: 9/15 at low, medium and high
every · x · 2026-09-29
Every's team benchmarked Sonnet 5.5's effort setting: Kieran Klaassen ran the same 15 coding tasks at low, medium and high effort — and passed 9 in every case. Cranking up reasoning didn't move the needle.
Design work tells a similar story: Tyler Nishida produced strong design pieces at medium effort, and while max/ultracode settings also delivered impressive results, his take is "you might as well reach for Opus 5.5" at that point.
Their practical guidance: use medium effort for work you can review and steer, and switch to Opus for longer, more detailed builds. Full Sonnet 5.5 Vibe Check linked in the post.
More from coding & agent
- Vibe-coded ComfyUI node fixes gibberish text in AI-generated speech bubbles — arturor1990 · 2026-09-29
- MetaLint: Qwen3-4B lifts code lint detection F-score 2.7x to 70.4%, matching o3-mini — dan_fried · 2026-09-29
- Building an evidence-driven incident response agent with Hindsight memory — CremeIntelligent1031 · 2026-09-29
- Hackathon prototype: incident agent that remembers failed fixes lifts accuracy 0% to 50% — Fabulous-Ad9781 · 2026-09-29
- Developer: AI agents cut my prototype time from two weeks to one day — pramodk73 · 2026-09-29
- gemini-cli PR enables /skill-name activation in non-interactive mode — Hariharanpugazh · 2026-09-29