开源大型语言模型
开源大型语言模型
Section titled “开源大型语言模型”Open Source LLMs tools for large language models.
Mistral (Large 3, Ministral, Small)
Section titled “Mistral (Large 3, Ministral, Small)”- Website: https://mistral.ai/
- Description: 开源模型:675B MoE 大型,小型多模态 Ministral,Apache 2.0(2025)。
- Pricing: 🆓
DeepSeek (V3.2, R1)
Section titled “DeepSeek (V3.2, R1)”- Website: https://deepseek.com/
- Description: 开源推理模型,匹配 GPT-5,MoE 架构(671B)。
- Pricing: 🆓
LLaMA 4 (Scout, Maverick)
Section titled “LLaMA 4 (Scout, Maverick)”- Website: https://ai.meta.com/llama/
- Description: Meta 的 MoE 模型:109B Scout(10M 上下文),400B Maverick 多模态。
- Pricing: 🆓
Qwen 3 (VL-235B)
Section titled “Qwen 3 (VL-235B)”- Website: https://qwenlm.github.io/
- Description: 阿里巴巴的万亿参数混合模型,支持视觉-语言功能。
- Pricing: 🆓
GLM-4.5
Section titled “GLM-4.5”- Website: https://github.com/THUDM/GLM-4
- Description: 智谱的130B模型,在LMArena排名第2,采用MIT许可证。
- Pricing: 🆓
Kimi K2
Section titled “Kimi K2”- Website: https://moonshot.cn/
- Description: Moonshot AI 的强大中文大型语言模型,具有开放研究。
- Pricing: 🆓
Gemma 3
Section titled “Gemma 3”- Website: https://ai.google.dev/gemma
- Description: 谷歌的轻量级开放模型,采用Gemma许可证。
- Pricing: 🆓
Falcon 3
Section titled “Falcon 3”- Website: https://falconllm.tii.ae/
- Description: TII 发布的最新开源大型语言模型,性能有所提升。
- Pricing: 🆓