{ "hf_id": "openbmb/MiniCPM5-2B", "meta": { "title": "MiniCPM5-2B", "slug": "minicpm5-2b", "provider": "MiniCPM (OpenBMB)", "description": "MiniCPM5-2B — dense 2B LLM with hybrid Think/No-Think reasoning, native 128K context, and native tool calling support, built on the standard Llama architecture", "date_added": "2026-09-07", "date_updated": "2026-09-07", "tasks": [ "text" ], "performance_headline": "2B-class open-source model with strong reasoning and tool use", "related_recipes": [ "openbmb/MiniCPM5-1B", "openbmb/MiniCPM-V-4.6" ] }, "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 openbmb/MiniCPM5-2B \\\n --tensor-parallel-size 1 \\\n --enable-auto-tool-choice \\\n --tool-call-parser minicpm5", "argv": [ "vllm", "serve", "openbmb/MiniCPM5-2B", "--tensor-parallel-size", "1", "--enable-auto-tool-choice", "--tool-call-parser", "minicpm5" ], "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 openbmb/MiniCPM5-2B \\\n --tensor-parallel-size 1 \\\n --enable-auto-tool-choice \\\n --tool-call-parser minicpm5", "docker_argv": [ "docker", "run", "--gpus", "all", "--privileged", "--ipc=host", "-p", "8000:8000", "-v", "~/.cache/huggingface:/root/.cache/huggingface", "vllm/vllm-openai:latest", "openbmb/MiniCPM5-2B", "--tensor-parallel-size", "1", "--enable-auto-tool-choice", "--tool-call-parser", "minicpm5" ], "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": { "multi_node_tp": "/openbmb/MiniCPM5-2B/strategies/multi_node_tp.json", "kv_store_centralized_mooncake": "/openbmb/MiniCPM5-2B/strategies/kv_store_centralized_mooncake.json", "kv_store_distributed_mooncake": "/openbmb/MiniCPM5-2B/strategies/kv_store_distributed_mooncake.json" }, "by_hardware": { "h200": "/openbmb/MiniCPM5-2B/hw/h200.json", "h100": "/openbmb/MiniCPM5-2B/hw/h100.json", "b200": "/openbmb/MiniCPM5-2B/hw/b200.json", "gb200": "/openbmb/MiniCPM5-2B/hw/gb200.json", "b300": "/openbmb/MiniCPM5-2B/hw/b300.json", "gb300": "/openbmb/MiniCPM5-2B/hw/gb300.json", "mi300x": "/openbmb/MiniCPM5-2B/hw/mi300x.json", "mi325x": "/openbmb/MiniCPM5-2B/hw/mi325x.json", "mi350x": "/openbmb/MiniCPM5-2B/hw/mi350x.json", "mi355x": "/openbmb/MiniCPM5-2B/hw/mi355x.json" } }, "model": { "model_id": "openbmb/MiniCPM5-2B", "min_vllm_version": "0.21.0", "architecture": "dense", "parameter_count": "2B", "active_parameters": "2B", "context_length": 131072, "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": { "reasoning": { "description": "Enable hybrid deep-thinking mode — the chat template wraps a block before the answer. Recommended sampling shifts to temperature=0.9, top_p=0.95. Leave off for fast No-Think responses (temperature=0.7).", "args": [ "--default-chat-template-kwargs", "{\"enable_thinking\": true}" ] }, "tool_calling": { "description": "Enable native tool calling with the MiniCPM5 XML parser (merged in vLLM PR #43175).", "args": [ "--enable-auto-tool-choice", "--tool-call-parser", "minicpm5" ] } }, "opt_in_features": [ "reasoning" ], "variants": { "default": { "precision": "bf16", "vram_minimum_gb": 5, "description": "BF16 weights (~2B params) — fits on a single consumer GPU" } }, "guide": "## Overview\n\n**MiniCPM5-2B** is the 2B checkpoint in OpenBMB's MiniCPM5 series — a dense\nmodel built for on-device and resource-constrained deployment, with strong\nperformance on agentic tool use, code generation, and reasoning tasks.\nIt uses the **standard `LlamaForCausalLM` architecture**, so vLLM loads it\nnatively with no custom kernels or model-code fork.\n\nIts headline feature is **hybrid reasoning**: a single checkpoint serves as\nboth a fast assistant (No-Think) and a deliberate reasoner (Think), toggled by\nthe chat template's `enable_thinking` flag.\n\n## Prerequisites\n\n- **vLLM ≥ 0.21.0** — MiniCPM5-2B is supported natively as of the v0.21.0\n release.\n\n## Launch command\n\nUse the command builder above. The baseline is simply:\n\n```bash\nvllm serve openbmb/MiniCPM5-2B --port 8000\n```\n\nAt ~2B params the model fits on a single GPU (TP=1). It supports the full\nnative 128K context; drop `--max-model-len` to `8192` / `32768` to free KV\ncache on small GPUs.\n\n## Reasoning modes\n\nToggle the **Reasoning** feature to serve with deep-thinking on by default\n(`--default-chat-template-kwargs '{\"enable_thinking\": true}'`). You can also\nflip it per request via `chat_template_kwargs`:\n\n```bash\ncurl http://localhost:8000/v1/chat/completions \\\n -H \"Content-Type: application/json\" \\\n -d '{\n \"model\": \"openbmb/MiniCPM5-2B\",\n \"messages\": [{\"role\": \"user\", \"content\": \"Explain GQA in one sentence.\"}],\n \"temperature\": 0.9, \"top_p\": 0.95, \"max_tokens\": 1024,\n \"chat_template_kwargs\": {\"enable_thinking\": true}\n }'\n```\n\n| Mode | `enable_thinking` | `temperature` | `top_p` |\n| --- | --- | --- | --- |\n| Think | `true` | 0.9 | 0.95 |\n| No-Think | `false` | 1.0 | 0.95 |\n\n## Tool calling\n\nMiniCPM5-2B emits **XML-style** tool calls. The `minicpm5` parser\n([PR #43175](https://github.com/vllm-project/vllm/pull/43175)) is natively\nsupported in vLLM. Enable it with:\n\n```bash\nvllm serve openbmb/MiniCPM5-2B --port 8000 \\\n --enable-auto-tool-choice \\\n --tool-call-parser minicpm5\n```\n\n## References\n\n- [Model card](https://huggingface.co/openbmb/MiniCPM5-2B)\n- [vLLM deployment cookbook](https://github.com/OpenBMB/MiniCPM/blob/main/docs/deployment/vllm.md)\n- [MiniCPM Tech Report](https://arxiv.org/pdf/2506.07900)\n", "source": { "source": "vllm", "repo": "vllm-project/vllm", "ref": "main", "sha": "564ac0de1b228568cafcbfbe8ccaccc82d6023eb", "path": "recipes/models/openbmb/MiniCPM5-2B.yaml", "edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/openbmb/MiniCPM5-2B.yaml", "overlays": [] } }