Jais 3 — Inception / MBZUAI Institute of Foundation Models 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
Jais 3 是 Inception 与 MBZUAI 基础模型研究院推出的阿拉伯语-英语模型系列，基于 K2 Horizon 架构：包括 0.9B 至 32B 的稠密模型和 375B 的混合专家模型，均支持 512K token 上下文，采用 Apache-2.0 许可。官方 GGUF 和 MLX 版本覆盖稠密模型。
变体
Jais 3 最小的变体在 Q4_K_M 下约需 6 GB 显存——量化权重加框架开销，不含 KV 缓存。
| 模型 | Q4 下显存 | 显存 | 上下文 | 运行 |
|---|---|---|---|---|
| Jais 3 7B → 7B | ~6.4 GB | 524,288 | ollama run jais-3 | |
| Jais 3 32B → 32B | ~21.9 GB | 524,288 | ollama run jais-3 | |
| Jais 3 375B-A23B → 375B (23B active) | ~229.7 GB | 524,288 | ollama run jais-3 |
显存为 Q4_K_M 下的量化权重加开销，与 GPU 与显存检测器使用同一引擎计算。
如何在本地运行 Jais 3
安装 Ollama，然后拉取标签。
ollama run jais-3在上方选择一个尺寸，查看它自己的显存、速度估算和安装命令。
许可证
Commercial use permitted. No usage restrictions beyond attribution.
适用于: Jais 3 7B, Jais 3 32B, Jais 3 375B-A23B推荐 GPU
目录中能在本地运行 Jais 3（至少 6 GB 显存）的最便宜 GPU 是 Intel Arc B570 (10 GB).
Jais 3 — 常见问题
How much VRAM does Jais 3 need?
Jais 3 needs about 6 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Jais 3 7B (6 GB, Q4_K_M); Jais 3 32B (22 GB, Q4_K_M); Jais 3 375B-A23B (230 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Jais 3 on an RTX 4090 (24 GB)?
Yes — Jais 3 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Jais 3?
Q4_K_M is the best balance of quality and VRAM for Jais 3 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Jais 3 with Ollama?
Jais 3 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.