gpt-oss-20b beats Opus on a task while being 1,000x cheaper and faster
scaling01 · x · 2026-08-04
The author says they ran into a task where gpt-oss-20b outperformed Opus by a huge margin.
Their claim is blunt: the smaller model was not just better on this task, but also around 1,000x cheaper, faster, and more reliable.
Related event: gpt-oss-20b Outperforms Opus in Real-World Tasks at 1/1000th Cost(2 posts)→
More from Models
- Meituan’s LongCat 2.0 is praised for domestic chips but judged mid-pack on quality — 葬AI · 2026-08-04
- Google lets Gemini 3.5 Flash and 3.6 Flash use Maps and Search together — OfficialLoganK · 2026-08-04
- DeepSeek V4 Flash tops the Vals Index above 60 at 35x lower cost — zephyr_z9 · 2026-08-04
- Vibe Code Bench shows V4 Flash beating GLM 5.2 on score and price — teortaxesTex · 2026-08-04
- Claude aced advanced math tutoring, then botched a flight-speed table — Recent-Day3062 · 2026-08-04
- Celeris-1 claims 2,038 tokens per second and fastest-on-leaderboard status — Scobleizer · 2026-08-04