Open-Weight Model Library
作者： Jakub Rusinowski · 最后更新： 2026年8月15日
Comprehensive library of 71 model families (154 variants) with VRAM requirements, quantization options, and one-click Ollama setup commands.
- BitNet b1.58 (Microsoft) — A research paradigm shift by Microsoft. BitNet b1.58 replaces 16-bit weights with ternary {-1, 0, 1}…
- DeepSeek R1 (DeepSeek) — DeepSeek R1 rivals proprietary top-tier models in reasoning and coding tasks using a Mixture-of-Expe…
- Llama 3.3 (Meta) — Meta's latest flagship model. Llama 3.3 70B delivers state-of-the-art performance, matching the mass…
- Llama 3.1 Family (Meta) — The reigning champion of open weights for consumer hardware. Llama 3.1 features a massive 128k conte…
- Nemotron 70B (NVIDIA) — NVIDIA's specialized fine-tune of Llama 3.1 70B. It tops many leaderboards for helpfulness and instr…
- Command R Family (Cohere) — Optimized specifically for RAG (Retrieval Augmented Generation) and tool use. Command R models excel…
- Phi-4 Family (Microsoft) — Microsoft's latest model focusing on high-quality synthetic data training. Phi-4 14B outperforms man…
- Phi 3.5 Family (Microsoft) — Microsoft's highly efficient small language models (SLMs). Perfect for mobile devices, edge computin…
- Qwen 2.5 Family (Alibaba Cloud) — The coding and math powerhouse of the open-weight world. Qwen 2.5 models consistently outperform Lla…
- Gemma 2 Family (Google) — Built from the same research and technology as Google Gemini. Gemma 2 uses knowledge distillation fo…
- Mistral Family (Mistral AI) — The European champion of open AI. Known for efficiency and the 'Nemo' collaborative model with NVIDI…
- Yi 1.5 Family (01.AI) — High-performance models from 01.AI. Known for strong bilingual (English/Chinese) capabilities and co…
- Granite 3.0 (IBM) — IBM's enterprise-grade model family. Granite 3.0 is Apache 2.0 licensed and rigorously trained for c…
- Llama 4 (Meta) — Meta's groundbreaking Mixture-of-Experts (MoE) series. Llama 4 uses a sparse MoE architecture where …
- Gemma 3 (Google) — Gemma 3 delivers exceptional quality-per-VRAM ratio, with the 4B model outperforming many 7B competi…
- Gemma 4 (Google) — Google's fourth-generation open model family. The first wave (E2B, E4B, 26B-A4B MoE, 31B dense) laun…
- Llama 3.2 Family (Meta) — Meta's multimodal and edge-optimized series. Llama 3.2 introduces vision capabilities (11B and 90B v…
- DeepSeek V3 (DeepSeek) — DeepSeek's groundbreaking non-reasoning frontier model. Trained for just $5.5M — 1/10th the cost of …
- Qwen 3 (Alibaba Cloud) — Alibaba's model series with a hybrid thinking mode. Qwen 3 models can toggle between fast response m…
- Qwen 2.5 VL (Alibaba Cloud) — Alibaba's powerful vision-language models that understand images, documents, charts, and videos alon…