{ "hf_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct", "meta": { "title": "Llama-4-Scout", "slug": "llama-4-scout", "provider": "Meta", "description": "Llama 4 Scout 17B-16E MoE model with NVIDIA FP8/FP4 variants, fits on a single GPU with quantization", "date_added": "2025-08-06", "date_updated": "2026-09-12", "tasks": [ "text" ], "performance_headline": "", "related_recipes": [ "meta-llama/Llama-3.3-70B-Instruct" ], "hardware": { "h100": "verified", "b200": "verified", "gb200": "verified", "mi300x": "verified", "mi325x": "verified", "mi355x": "verified", "xeon6": "verified" } }, "recommended_command": { "hardware": "b200", "strategy": "single_node_tp", "variant": "default", "node_count": 1, "deploy_type": "single_node", "env": {}, "docker_image": "vllm/vllm-openai:latest", "command": "vllm serve meta-llama/Llama-4-Scout-17B-16E-Instruct \\\n --tensor-parallel-size 8 \\\n --async-scheduling \\\n --no-enable-prefix-caching \\\n --max-num-batched-tokens 8192 \\\n --compilation-config '{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}'", "argv": [ "vllm", "serve", "meta-llama/Llama-4-Scout-17B-16E-Instruct", "--tensor-parallel-size", "8", "--async-scheduling", "--no-enable-prefix-caching", "--max-num-batched-tokens", "8192", "--compilation-config", "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}" ], "docker_command": "docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n vllm/vllm-openai:latest meta-llama/Llama-4-Scout-17B-16E-Instruct \\\n --tensor-parallel-size 8 \\\n --async-scheduling \\\n --no-enable-prefix-caching \\\n --max-num-batched-tokens 8192 \\\n --compilation-config '{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}'", "docker_argv": [ "docker", "run", "--gpus", "all", "--privileged", "--ipc=host", "-p", "8000:8000", "-v", "~/.cache/huggingface:/root/.cache/huggingface", "vllm/vllm-openai:latest", "meta-llama/Llama-4-Scout-17B-16E-Instruct", "--tensor-parallel-size", "8", "--async-scheduling", "--no-enable-prefix-caching", "--max-num-batched-tokens", "8192", "--compilation-config", "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}" ], "strategy_spec": { "name": "single_node_tp", "deploy_type": "single_node", "display_name": "Tensor Parallel", "short_label": "TP", "orientation": "latency", "description": "Single-node tensor parallel. Splits the model across all local GPUs. TP size is set to the GPU count at deploy time. The simplest multi-GPU strategy — works for all model architectures.\n", "hardware_match": { "min_gpus": 1, "max_gpus": 8, "multi_node": false }, "vllm_args": [], "parallel_flag": "--tensor-parallel-size" }, "hardware_profile": { "brand": "NVIDIA", "generation": "blackwell", "compute_arch": "sm100", "display_name": "B200", "description": "NVIDIA B200 SXM 180 GB HBM3e · 8-GPU HGX B200 node", "gpu_count": 8, "vram_gb": 1440, "multi_node": false }, "alternatives": { "single_node_tep": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/single_node_tep.json", "single_node_dep": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/single_node_dep.json", "multi_node_tp": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/multi_node_tp.json", "multi_node_tp_pp": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/multi_node_tp_pp.json", "multi_node_tep": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/multi_node_tep.json", "multi_node_dep": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/multi_node_dep.json", "kv_store_centralized_mooncake": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/kv_store_centralized_mooncake.json", "kv_store_distributed_mooncake": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/strategies/kv_store_distributed_mooncake.json" }, "by_hardware": { "b200": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/b200.json", "h100": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/h100.json", "gb200": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/gb200.json", "mi300x": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/mi300x.json", "mi325x": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/mi325x.json", "mi355x": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/mi355x.json", "xeon6": "/meta-llama/Llama-4-Scout-17B-16E-Instruct/hw/xeon6.json" } }, "model": { "model_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct", "min_vllm_version": "0.12.0", "architecture": "moe", "parameter_count": "109B", "active_parameters": "17B", "context_length": 10485760, "base_args": [], "base_env": {}, "install": { "pip": { "command": "uv venv\nsource .venv/bin/activate\nuv pip install -U vllm --torch-backend auto" }, "docker": { "command": "docker pull vllm/vllm-openai:latest" } } }, "features": {}, "opt_in_features": [], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 262, "description": "Full precision BF16", "supported_hardware": [ "h100", "b200", "gb200", "mi300x", "mi325x", "mi355x", "xeon6" ] }, "fp8": { "model_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8", "precision": "fp8", "vram_minimum_gb": 131, "description": "NVIDIA FP8 quantization for Hopper and Blackwell, fits on 1x H100", "supported_hardware": [ "h100", "b200", "gb200" ], "extra_args": [ "--kv-cache-dtype", "fp8" ], "extra_env": {}, "json": "/nvidia/Llama-4-Scout-17B-16E-Instruct-FP8.json" }, "nvfp4": { "model_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4", "precision": "nvfp4", "vram_minimum_gb": 65, "description": "NVIDIA NVFP4 quantized weights for Blackwell GPUs", "supported_hardware": [ "b200", "gb200" ], "extra_args": [ "--kv-cache-dtype", "fp8" ], "json": "/nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4.json" }, "redhat_fp8": { "model_id": "RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic", "precision": "fp8", "label": "FP8 (RedHat)", "vram_minimum_gb": 131, "description": "RedHatAI FP8 (E4M3) with dynamic per-token activation quantization", "supported_hardware": [ "h100", "b200", "gb200" ], "extra_args": [ "--kv-cache-dtype", "fp8" ], "json": "/RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic.json" }, "redhat_nvfp4": { "model_id": "RedHatAI/Llama-4-Scout-17B-16E-Instruct-NVFP4", "precision": "nvfp4", "label": "NVFP4 (RedHat)", "vram_minimum_gb": 65, "description": "RedHatAI compressed-tensors NVFP4 quantization for Blackwell GPUs", "supported_hardware": [ "b200", "gb200" ], "extra_args": [ "--kv-cache-dtype", "fp8" ], "extra_env": { "VLLM_USE_FLASHINFER_MOE_FP4": "1" }, "json": "/RedHatAI/Llama-4-Scout-17B-16E-Instruct-NVFP4.json" }, "redhat_fp8_block": { "model_id": "RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-block", "precision": "fp8", "label": "FP8 Block (RedHat)", "vram_minimum_gb": 131, "description": "RedHatAI FP8 block quantization", "supported_hardware": [ "h100", "b200", "gb200" ], "extra_args": [ "--kv-cache-dtype", "fp8" ], "json": "/RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-block.json" } }, "hardware_overrides": { "cpu": { "extra_args": [ "--max-model-len", "8192" ] }, "hopper": { "extra_args": [ "--async-scheduling", "--no-enable-prefix-caching", "--max-num-batched-tokens", "8192" ], "extra_env": {} }, "blackwell": { "extra_args": [ "--async-scheduling", "--no-enable-prefix-caching", "--max-num-batched-tokens", "8192", "--compilation-config", "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}" ] }, "amd": { "extra_args": [ "--no-enable-prefix-caching", "--max-num-batched-tokens", "16384", "--max-num-seqs", "64", "--max-model-len", "32000" ], "extra_env": { "VLLM_ROCM_USE_AITER": "1" } } }, "guide": "## Overview\n\nLlama 4 Scout is Meta's MoE model with 17B active parameters across 16 experts\n(109B total). NVIDIA provides FP8 and FP4 quantized variants. With FP4 quantization,\nthe model fits on a single B200 GPU — making it one of the most accessible MoE models.\n\n## Prerequisites\n\n- Hardware: 1x B200 (FP4), 1x H100 (FP8), 4x GPUs (BF16), 8x MI300X/MI325X/MI355X, or 4x Xeon6/Xeon5 NUMA nodes\n- vLLM >= 0.12.0\n- CUDA Driver >= 575 for GPUs\n- Docker with NVIDIA Container Toolkit (recommended) for GPUs\n- License: Must agree to Meta's Llama 4 Scout Community License for GPUs\n\n### AMD ROCm\n\nThe ROCm wheel requires Python 3.12, ROCm 7.0+, and glibc >= 2.35. Use the\nROCm Docker image when your host environment does not meet those requirements.\n\n```bash\nuv venv --python 3.12\nsource .venv/bin/activate\nuv pip install vllm --extra-index-url https://wheels.vllm.ai/rocm/\n```\n\nLaunch on 8x MI300X / MI325X / MI355X with AITER enabled:\n\n```bash\nexport VLLM_ROCM_USE_AITER=1\nvllm serve meta-llama/Llama-4-Scout-17B-16E-Instruct \\\n --tensor-parallel-size 8 \\\n --max-num-seqs 64 \\\n --no-enable-prefix-caching \\\n --max-num-batched-tokens 16384 \\\n --max-model-len 32000\n```\n\nThe first AITER launch may take several minutes while ROCm kernels are compiled and cached.\n\n### pip (Intel Xeon 6 CPUs)\nFor Intel and AMD x86 CPUs, follow the [CPU pre-built wheels](https://docs.vllm.ai/en/latest/getting_started/installation/cpu/#pre-built-wheels) installation instructions.\n\n### Docker (Intel Xeon 6 CPUs)\n\n```bash\ndocker pull vllm/vllm-openai-cpu:latest-x86_64 # For Intel Xeon 6\n```\n\n### Intel Xeon 6 Deployment via Docker\n\nLaunch the x86 CPU vLLM Docker container for `meta-llama/Llama-4-Scout-17B-16E-Instruct`:\n\n```bash\ndocker run \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n vllm/vllm-openai-cpu:latest-x86_64 meta-llama/Llama-4-Scout-17B-16E-Instruct \\\n --tensor-parallel-size 4 \\\n --max-model-len 8192\n```\n\n`--max-model-len 8192` is retained because it is the model-specific row argument\nin the validation workbook. TP/DP and host-specific CPU placement are intentionally\nnot prescribed.\n\nThe following validation settings are not model requirements and are intentionally\nnot portable CPU recipe defaults:\n\n- `--max-num-batched-tokens` and `--max-num-seqs`: scheduler batch/concurrency\n tuning for workload shape, latency targets, and platform capacity.\n- `--gpu-memory-utilization`: platform memory-budget tuning.\n- `--no-enable-prefix-caching`: workload/benchmark cache-policy tuning.\n- `VLLM_ENGINE_ITERATION_TIMEOUT_S`: runtime operational timeout tuning.\n\nHardware-specific overrides may still use these settings when they are part of a\nvalidated platform configuration.\n\n## Client Usage\n\n```python\nfrom openai import OpenAI\n\nclient = OpenAI(base_url=\"http://localhost:8000/v1\", api_key=\"unused\")\nresponse = client.chat.completions.create(\n model=\"nvidia/Llama-4-Scout-17B-16E-Instruct-FP8\",\n messages=[{\"role\": \"user\", \"content\": \"Explain MoE models briefly.\"}],\n)\nprint(response.choices[0].message.content)\n```\n\n## Troubleshooting\n\n**FP4 only works on Blackwell:**\nFP4 quantization requires compute capability 10.0 (B200/GB200). Use FP8 on Hopper.\n\n**TP=1 recommended for best throughput:**\nFor maximum throughput per GPU, keep TP=1. Increase TP to 2/4/8 for lower latency.\n\n## References\n\n- [Llama-4-Scout-17B-16E-Instruct](https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct)\n- [NVIDIA FP8 variant](https://huggingface.co/nvidia/Llama-4-Scout-17B-16E-Instruct-FP8)\n- [NVIDIA FP4 variant](https://huggingface.co/nvidia/Llama-4-Scout-17B-16E-Instruct-FP4)\n", "source": { "source": "vllm", "repo": "vllm-project/vllm", "ref": "main", "sha": "564ac0de1b228568cafcbfbe8ccaccc82d6023eb", "path": "recipes/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.yaml", "edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/meta-llama/Llama-4-Scout-17B-16E-Instruct.yaml", "overlays": [] } }