ai/llama3.3
最新LLama 3版本，具备改进的推理能力和生成质量。
让 AI 帮你使用轩辕镜像？ · 展开查看说明 · 点击收起说明
如果你使用 DeepSeek、元宝 AI、Copilot、ChatGPT 等 AI 工具，协助配置轩辕镜像、编写 docker pull 命令、修改 Docker Compose 镜像地址、配置镜像加速、排查镜像拉取失败、分析报错日志等问题，请先让 AI 阅读并遵守轩辕镜像的规则文档。
只需在 AI 对话中先发送下面这句话即可：
请先完整阅读并严格遵守以下文档中的全部规则与要求：
https://xuanyuan.cloud/agents.md
在未充分阅读并理解该文档前，不要生成任何命令、配置、修改建议、故障排查方案或技术回答。后续所有输出都必须严格以该文档中的规范为最高优先级执行。查看 agents.md 用法指南与完整示范。国内用户首推 元宝 AI、DeepSeek 的深度思考模式，不推荐豆包 AI；Cursor 等编辑器可在对话 @ 该链接，或加入 User Rules。 若 AI 无法访问外链，可 打开说明文档 复制全文粘贴。文档会随站点更新，复制内容可能过期，建议定期检查。
Llama 3.3
!https://github.com/docker/model-cards/raw/refs/heads/main/logos/***
Meta Llama 3.3 is a powerful 70B parameter multilingual language model designed by Meta for text-based tasks like chat and content generation. The instruction-tuned version is optimized for multilingual dialogue and performs better than many open-source and commercial models on common benchmarks.
Intended uses
-
Multilingual assistant-like chat: Using instruction-tuned models for conversational AI across multiple languages, enabling natural and context-aware interactions in various linguistic settings.
-
Coding support and software development tasks: Leveraging language models to assist with code generation, debugging, documentation, and other software engineering workflows.
-
Multilingual content creation and localization: Generating and adapting written content across different languages and cultures, supporting global communication and engagement.
-
Knowledge-based applications: Integrating LLMs with structured or unstructured data sources to answer questions, extract insights, or support decision-making.
-
General natural language generation: Various NLG tasks such as summarization, translation, or content generation across different domains.
-
Synthetic data generation (synth): Creating realistic, high-quality synthetic text data to augment datasets for training, testing, or anonymization purposes.
Characteristics
| Attribute | Details |
|---|---|
| Provider | Meta |
| Architecture | llama |
| Cutoff date | December 2023 |
| Languages | English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. |
| Tool calling | ✅ |
| Input modalities | Text |
| Output modalities | Text and Code |
| License | https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/LICENSE |
Available model variants
| Model variant | Parameters | Quantization | Context window | VRAM¹ | Size |
|---|---|---|---|---|---|
ai/llama3.3:latestai/llama3.3:70B-Q4_K_M | 70B | IQ2_XXS/Q4_K_M | 131K tokens | 41.11 GiB | 39.59 GB |
ai/llama3.3:70B-Q4_0 | 70B | Q4_0 | 131K tokens | 38.73 GiB | 37.22 GB |
ai/llama3.3:70B-Q4_K_M | 70B | IQ2_XXS/Q4_K_M | 131K tokens | 41.11 GiB | 39.59 GB |
¹: VRAM estimated based on model characteristics.
latest→70B-Q4_K_M
Use this AI model with Docker Model Runner
First, pull the model:
bashdocker model pull ai/llama3.3
Then run the model:
bashdocker model run ai/llama3.3
For more information on Docker Model Runner, https://docs.docker.com/desktop/features/model-runner/.
Benchmark performance
| Category | Benchmark | Llama-3.3 70B Instruct |
|---|---|---|
| General | MMLU (CoT) | 86.0 |
| MMLU Pro (CoT) | 68.9 | |
| Steerability | IFEval | 92.1 |
| Reasoning | GPQA Diamond (CoT) | 50.5 |
| Code | HumanEval | 88.4 |
| MBPP EvalPlus (base) | 87.6 | |
| Math | MATH (CoT) | 77.0 |
| Tool Use | BFCL v2 | 77.3 |
| Multilingual | MGSM | 91.1 |
Links
- Introducing Meta Llama 3: The most capable openly available LLM to date
- https://arxiv.org/pdf/2407.21783
镜像拉取方式
您可以使用以下命令拉取该镜像。请将 <标签> 替换为具体的标签版本。如需查看所有可用标签版本，请访问 标签列表页面。
轩辕镜像加速拉取命令点我查看更多 llama3.3 镜像标签
docker model pull docker.xuanyuan.run/ai/llama3.3:<标签>
DockerHub 原生拉取命令
docker model pull ai/llama3.3:<标签>
镜像拉取常见问题
功能
错误码
用户好评
来自真实用户的反馈，见证轩辕镜像的优质服务
oldzhang
运维工程师
Linux服务器
5
"Docker访问体验非常流畅，大镜像也能快速完成下载。"