Qwen 3.8 Flash beats Opus on SWE-bench Pro
Hesamation · x · 2026-08-26
Qwen 3.8 Flash is praised for performance:
- 125B MoE with 6B active params.
- SWE-bench Pro: 62.5 vs Opus 4.6 Max at 53.4.
- Qwen 4 architecture preview.
- 9x cheaper to train than Qwen3.7-Plus.
More from Models
- Qwen3.8-Flash-Next appears early on Hugging Face — alex_bit_ · 2026-08-26
- Qwen 3.8-Max scores 61.7 on SWE-Bench Pro, beating Opus 4.6's 53.4 — close_Meal6005 · 2026-08-26
- Grok 4.6 Wins Mario Clone Coding Challenge Against Ox Alpha, Gemini, and Kimi — CodeByPoonam · 2026-08-26
- Heavy Users Notice ChatGPT's Temporal Context Hallucinations — DV_Studio_Dev · 2026-08-26
- Tip: opencode's new model is reportedly a nerfed multimodal GLM-5.3 — PawelHuryn · 2026-08-26
- DeepSeek's AI Demonstrates Self-Upgrading Capabilities with New Harness Tool — Two Minute Papers · 2026-08-26