{ "hf_id": "poolside/Laguna-M.1", "meta": { "title": "Laguna M.1", "slug": "laguna-m.1", "provider": "Poolside", "description": "Poolside's 225B total / 23B activated MoE coding model with global attention, native interleaved reasoning, and 256K context — designed for agentic coding and long-horizon work.", "date_added": "2026-06-18", "date_updated": "2026-06-18", "tasks": [ "text" ], "performance_headline": "225B/23B-A MoE for agentic coding — 74.6% SWE-bench Verified", "related_recipes": [ "poolside/Laguna-XS.2" ] }, "recommended_command": { "hardware": "h200", "strategy": "single_node_tp", "variant": "default", "node_count": 1, "deploy_type": "single_node", "env": {}, "docker_image": "vllm/vllm-openai:latest", "command": "vllm serve poolside/Laguna-M.1 \\\n --tensor-parallel-size 8 \\\n --enable-auto-tool-choice \\\n --tool-call-parser poolside_v1 \\\n --reasoning-parser poolside_v1 \\\n --default-chat-template-kwargs '{\"enable_thinking\": true}'", "argv": [ "vllm", "serve", "poolside/Laguna-M.1", "--tensor-parallel-size", "8", "--enable-auto-tool-choice", "--tool-call-parser", "poolside_v1", "--reasoning-parser", "poolside_v1", "--default-chat-template-kwargs", "{\"enable_thinking\": 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 poolside/Laguna-M.1 \\\n --tensor-parallel-size 8 \\\n --enable-auto-tool-choice \\\n --tool-call-parser poolside_v1 \\\n --reasoning-parser poolside_v1 \\\n --default-chat-template-kwargs '{\"enable_thinking\": true}'", "docker_argv": [ "docker", "run", "--gpus", "all", "--privileged", "--ipc=host", "-p", "8000:8000", "-v", "~/.cache/huggingface:/root/.cache/huggingface", "vllm/vllm-openai:latest", "poolside/Laguna-M.1", "--tensor-parallel-size", "8", "--enable-auto-tool-choice", "--tool-call-parser", "poolside_v1", "--reasoning-parser", "poolside_v1", "--default-chat-template-kwargs", "{\"enable_thinking\": 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": "hopper", "compute_arch": "sm90", "display_name": "H200", "description": "NVIDIA H200 SXM 141 GB HBM3e · 8-GPU HGX node", "gpu_count": 8, "vram_gb": 1128, "multi_node": false }, "alternatives": { "single_node_tep": "/poolside/Laguna-M.1/strategies/single_node_tep.json", "single_node_dep": "/poolside/Laguna-M.1/strategies/single_node_dep.json", "multi_node_tp": "/poolside/Laguna-M.1/strategies/multi_node_tp.json", "multi_node_dep": "/poolside/Laguna-M.1/strategies/multi_node_dep.json", "multi_node_tep": "/poolside/Laguna-M.1/strategies/multi_node_tep.json", "kv_store_centralized_mooncake": "/poolside/Laguna-M.1/strategies/kv_store_centralized_mooncake.json", "kv_store_distributed_mooncake": "/poolside/Laguna-M.1/strategies/kv_store_distributed_mooncake.json" }, "by_hardware": { "h200": "/poolside/Laguna-M.1/hw/h200.json", "h100": "/poolside/Laguna-M.1/hw/h100.json", "b200": "/poolside/Laguna-M.1/hw/b200.json", "gb200": "/poolside/Laguna-M.1/hw/gb200.json", "b300": "/poolside/Laguna-M.1/hw/b300.json", "gb300": "/poolside/Laguna-M.1/hw/gb300.json", "mi300x": "/poolside/Laguna-M.1/hw/mi300x.json", "mi325x": "/poolside/Laguna-M.1/hw/mi325x.json", "mi350x": "/poolside/Laguna-M.1/hw/mi350x.json", "mi355x": "/poolside/Laguna-M.1/hw/mi355x.json" } }, "model": { "model_id": "poolside/Laguna-M.1", "min_vllm_version": "0.21.0", "architecture": "moe", "parameter_count": "225B", "active_parameters": "23B", "context_length": 262144, "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": { "tool_calling": { "description": "Enable automatic tool choice with Poolside's tool-call parser", "args": [ "--enable-auto-tool-choice", "--tool-call-parser", "poolside_v1" ] }, "reasoning": { "description": "Enable interleaved thinking (on by default) with Poolside's reasoning parser", "args": [ "--reasoning-parser", "poolside_v1", "--default-chat-template-kwargs", "{\"enable_thinking\": true}" ] } }, "opt_in_features": [], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 540, "description": "BF16 weights (225B) — fits on a single 8×H200/B200 node" }, "fp8": { "model_id": "poolside/Laguna-M.1-FP8", "precision": "fp8", "vram_minimum_gb": 270, "description": "Block-FP8 weights + FP8 KV cache — halves VRAM, runs on a single 4×H200/B200 node", "json": "/poolside/Laguna-M.1-FP8.json" }, "nvfp4": { "model_id": "poolside/Laguna-M.1-NVFP4", "precision": "nvfp4", "vram_minimum_gb": 135, "description": "NVFP4 weights (Blackwell only) — fits within ~135GB; pairs with FP8 KV cache", "json": "/poolside/Laguna-M.1-NVFP4.json" } }, "hardware_overrides": {}, "guide": "## Overview\n\n[Laguna M.1](https://huggingface.co/poolside/Laguna-M.1) is Poolside's 225B-total / 23B-activated Mixture-of-Experts model purpose-built for agentic coding and long-horizon work. It is competitive with state-of-the-art open-weight and frontier models on SWE-bench Verified (74.6%), SWE-bench Multilingual, SWE-Bench Pro, and Terminal-Bench 2.0.\n\n### Key features\n- **Large sparse MoE**: 70-layer transformer — the first 3 layers are dense SwiGLU, the remaining 67 are sparse MoE with **256 experts + 1 shared expert** and top-k=16 routing with auxiliary-loss-free load balancing.\n- **Global attention**: global attention across all layers with 64 Q-heads, 8 KV-heads, head dim 128, and softplus attention output gating.\n- **256K context**: RoPE with YaRN extends the context window to 262,144 tokens.\n- **Native interleaved reasoning**: thinking blocks emitted before and between tool calls, with *preserved thinking* (returning prior `reasoning` in history) recommended for best agentic performance. Toggled per-request via `enable_thinking`.\n- **Tool calling**: Poolside-specific tool-call protocol, parsed via `poolside_v1`.\n\n## Prerequisites\n\nLaguna M.1 support landed in vLLM via [PR #41129](https://github.com/vllm-project/vllm/pull/41129) (shared implementation with [Laguna XS.2](https://huggingface.co/poolside/Laguna-XS.2)) and ships in **vLLM 0.21.0 and later**.\n\n```bash\nuv pip install -U 'vllm>=0.21.0'\n```\n\nFP8 and NVFP4 quantized checkpoints are available at [Laguna-M.1-FP8](https://huggingface.co/poolside/Laguna-M.1-FP8) and [Laguna-M.1-NVFP4](https://huggingface.co/poolside/Laguna-M.1-NVFP4). Quantization is detected automatically from the checkpoint's `quantization_config`, so the launch command is identical — just swap the model ID (select the variant above).\n\n## Launch command\n\n### Single node (8×H200, BF16)\n```bash\nvllm serve poolside/Laguna-M.1 \\\n --tensor-parallel-size 8 \\\n --served-model-name laguna \\\n --enable-auto-tool-choice \\\n --tool-call-parser poolside_v1 \\\n --reasoning-parser poolside_v1 \\\n --default-chat-template-kwargs '{\"enable_thinking\": true}'\n```\n\nThe BF16 checkpoint (~450 GB of weights) fits on a single 8×H200 node. Use the **FP8** variant to run on 4 GPUs, or the **NVFP4** variant (Blackwell only) to fit within ~135 GB.\n\n## Controlling reasoning\n\nReasoning is enabled at the server when you pass `--default-chat-template-kwargs '{\"enable_thinking\": true}'` (included above). Poolside recommends **preserved thinking** for agentic coding — return the previous assistant turn's `reasoning` in the message history.\n\n```python\nfrom openai import OpenAI\nclient = OpenAI(base_url=\"http://localhost:8000/v1\", api_key=\"EMPTY\")\n\nresp = client.chat.completions.create(\n model=\"laguna\",\n messages=[{\"role\": \"user\", \"content\": \"Write a Python retry wrapper with exponential backoff.\"}],\n temperature=1.0,\n extra_query={\"top_k\": 20},\n stream=True,\n)\nfor chunk in resp:\n delta = chunk.choices[0].delta\n rc = delta.reasoning\n if rc:\n print(rc, end=\"\")\n if delta.content:\n print(delta.content, end=\"\")\n```\n\nRecommended sampling (matching Poolside's benchmark runs): **temperature=1.0, top_k=20**, thinking enabled.\n\nTo disable thinking, omit `--default-chat-template-kwargs` at launch, or pass `extra_body={\"chat_template_kwargs\": {\"enable_thinking\": False}}` per request.\n\n## References\n\n- [Model card](https://huggingface.co/poolside/Laguna-M.1)\n- [Release blog post](https://poolside.ai/blog/laguna-a-deeper-dive)\n- [Technical report](https://poolside.ai/assets/laguna/laguna-m1-xs2-technical-report.pdf)\n- [vLLM support PR #41129](https://github.com/vllm-project/vllm/pull/41129)\n", "source": { "source": "vllm", "repo": "vllm-project/vllm", "ref": "main", "sha": "389ffebe555705c85337025d5df108d0384d365e", "path": "recipes/models/poolside/Laguna-M.1.yaml", "edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/poolside/Laguna-M.1.yaml", "overlays": [] } }