1.3B参数就能在多项基准上超越更大模型,做边缘部署或移动端视觉应用的开发者值得一试,成本低效果强。
面壁智能发布MiniCPM-V 4.6,仅1.3B参数的视觉模型,在多项基准测试中表现强劲,甚至超越更大模型。该模型采用LLaVA-UHD v4技术,将视觉编码成本降低55%,适合消费级和移动硬件部署。在关键多模态和Artificial Analysis基准上,它超越了Gemma4-E2B-it和Qwen3.5-0.8B,且仅用Qwen3.5-0.8B 2.5%的token预算。在高分辨率图像处理中,TTFT(75.7ms)比Qwen3.5-0.8B快2.2倍,单张RTX 4090上吞吐量提升约1.5倍。模型已在Hugging Face、GitHub和ModelScope开源。
前几天在Huggingface看到模型论文了。 面壁智能的MiniCPM-V 4.6 ,竟然只有1.3B的视觉模型。 看Benchmark效果有点强,抽空测试下。
前几天在Huggingface看到模型论文了。 面壁智能的MiniCPM-V 4.6 ,竟然只有1.3B的视觉模型。 看Benchmark效果有点强,抽空测试下。 OpenBMB @OpenBMB 1/5 MiniCPM-V 4.6 (1.3B) is now live 🚀🚀 High-res visual processing, optimized for consumer-grade and mobile hardware. We’ve leveraged the latest LLaVA-UHD v4 technique to cut vision encoding costs by 55%, enabling native edge deployment with extreme efficiency. 🔥 Beats Gemma4-E2B-it and Qwen3.5-0.8B across key multimodal and Artificial Analysis benchmarks — scoring higher than Qwen3.5-0.8B using just 2.5% of its token budget. ⚡ TTFT (75.7ms) 2.2x Faster than Qwen3.5-0.8B even with 3136² high-res images. 🏗️ ~1.5x Token Throughput compared with Qwen3.5-0.8B on a single RTX 4090. Try the model here: 🤗 Hugging F huggingface.co/openbmb/MiniCP… MSBwc 💻 Gi github.com/OpenBMB/MiniCP… pa52tn 🔭 Model modelscope.cn/models/OpenBMB… lKPLbvK 🌐 We huggingface.co/spaces/openbmb… UrtD0YzM 📱 A github.com/OpenBMB/MiniCP… L7IOhm6zv Your browser does not support the video tag. 🔗 View on Twitter 🔗 View Quoted Tweet 💬 1 🔄 1 ❤️ 5 👀 4614 📊 3 ⚡