👀 You don't need a closed model to beat Qwen3.8-27B at agent coding. Nanbeige4… · @QwenAIFans · TwKit

@QwenAIFans

👀 You don't need a closed model to beat Qwen3.8-27B at agent coding. Nanbeige4.3-33B just dropped (Apache 2.0, BOSS Zhipin lab). Their HF table vs Qwen3.8-27B: • DeepSWE: 65.5 vs 42.5 • Terminal-Bench 2.1: 80.9 vs 73.0 • 512K context Caveat: vendor-reported. Qwen still leads SWE-Bench Pro 61.7–60.3. via Hugging Face 👇 https://huggingface.co/Nanbeige/Nanbeige4.3-33B

曝光 1435 · 评论 1 · 点赞 19 · 书签 3