{
"hf_id": "inclusionAI/Ring-2.6-1T",
"meta": {
"title": "Ring-2.6-1T",
"slug": "ring-2.6-1t",
"provider": "inclusionAI",
"description": "Ring-2.6-1T (BailingMoeV2_5) FP8 thinking model with 1T total / 50B active params, hybrid linear + MLA attention, 128K context",
"date_added": "2026-05-14",
"date_updated": "2026-05-15",
"tasks": [
"text"
],
"related_recipes": [
"inclusionAI/Ling-2.6-1T",
"inclusionAI/Ring-1T-FP8"
]
},
"recommended_command": {
"hardware": "h200",
"strategy": "multi_node_tp",
"variant": "default",
"node_count": 2,
"deploy_type": "multi_node",
"env": {
"GLOO_SOCKET_IFNAME": "$IFACE_NAME",
"NCCL_SOCKET_IFNAME": "$IFACE_NAME"
},
"docker_image": "vllm/vllm-openai:latest",
"head_command": "vllm serve inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 16 \\\n --nnodes 2 \\\n --node-rank 0 \\\n --master-addr $HEAD_IP",
"worker_command": "vllm serve inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 16 \\\n --nnodes 2 \\\n --node-rank 1 \\\n --master-addr $HEAD_IP \\\n --headless",
"worker_commands": [
"vllm serve inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 16 \\\n --nnodes 2 \\\n --node-rank 1 \\\n --master-addr $HEAD_IP \\\n --headless"
],
"head_argv": [
"vllm",
"serve",
"inclusionAI/Ring-2.6-1T",
"--trust-remote-code",
"--tensor-parallel-size",
"16",
"--nnodes",
"2",
"--node-rank",
"0",
"--master-addr",
"$HEAD_IP"
],
"worker_argv": [
"vllm",
"serve",
"inclusionAI/Ring-2.6-1T",
"--trust-remote-code",
"--tensor-parallel-size",
"16",
"--nnodes",
"2",
"--node-rank",
"1",
"--master-addr",
"$HEAD_IP",
"--headless"
],
"worker_argvs": [
[
"vllm",
"serve",
"inclusionAI/Ring-2.6-1T",
"--trust-remote-code",
"--tensor-parallel-size",
"16",
"--nnodes",
"2",
"--node-rank",
"1",
"--master-addr",
"$HEAD_IP",
"--headless"
]
],
"head_docker_command": "docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n -e GLOO_SOCKET_IFNAME=$IFACE_NAME \\\n -e NCCL_SOCKET_IFNAME=$IFACE_NAME \\\n vllm/vllm-openai:latest inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 16 \\\n --nnodes 2 \\\n --node-rank 0 \\\n --master-addr $HEAD_IP",
"worker_docker_command": "docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n -e GLOO_SOCKET_IFNAME=$IFACE_NAME \\\n -e NCCL_SOCKET_IFNAME=$IFACE_NAME \\\n vllm/vllm-openai:latest inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 16 \\\n --nnodes 2 \\\n --node-rank 1 \\\n --master-addr $HEAD_IP \\\n --headless",
"worker_docker_commands": [
"docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n -e GLOO_SOCKET_IFNAME=$IFACE_NAME \\\n -e NCCL_SOCKET_IFNAME=$IFACE_NAME \\\n vllm/vllm-openai:latest inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 16 \\\n --nnodes 2 \\\n --node-rank 1 \\\n --master-addr $HEAD_IP \\\n --headless"
],
"head_docker_argv": [
"docker",
"run",
"--gpus",
"all",
"--privileged",
"--ipc=host",
"-p",
"8000:8000",
"-v",
"~/.cache/huggingface:/root/.cache/huggingface",
"-e",
"GLOO_SOCKET_IFNAME=$IFACE_NAME",
"-e",
"NCCL_SOCKET_IFNAME=$IFACE_NAME",
"vllm/vllm-openai:latest",
"inclusionAI/Ring-2.6-1T",
"--trust-remote-code",
"--tensor-parallel-size",
"16",
"--nnodes",
"2",
"--node-rank",
"0",
"--master-addr",
"$HEAD_IP"
],
"worker_docker_argv": [
"docker",
"run",
"--gpus",
"all",
"--privileged",
"--ipc=host",
"-p",
"8000:8000",
"-v",
"~/.cache/huggingface:/root/.cache/huggingface",
"-e",
"GLOO_SOCKET_IFNAME=$IFACE_NAME",
"-e",
"NCCL_SOCKET_IFNAME=$IFACE_NAME",
"vllm/vllm-openai:latest",
"inclusionAI/Ring-2.6-1T",
"--trust-remote-code",
"--tensor-parallel-size",
"16",
"--nnodes",
"2",
"--node-rank",
"1",
"--master-addr",
"$HEAD_IP",
"--headless"
],
"worker_docker_argvs": [
[
"docker",
"run",
"--gpus",
"all",
"--privileged",
"--ipc=host",
"-p",
"8000:8000",
"-v",
"~/.cache/huggingface:/root/.cache/huggingface",
"-e",
"GLOO_SOCKET_IFNAME=$IFACE_NAME",
"-e",
"NCCL_SOCKET_IFNAME=$IFACE_NAME",
"vllm/vllm-openai:latest",
"inclusionAI/Ring-2.6-1T",
"--trust-remote-code",
"--tensor-parallel-size",
"16",
"--nnodes",
"2",
"--node-rank",
"1",
"--master-addr",
"$HEAD_IP",
"--headless"
]
],
"strategy_spec": {
"name": "multi_node_tp",
"deploy_type": "multi_node",
"display_name": "Multi-Node Tensor Parallel",
"short_label": "Multi-node TP",
"orientation": "latency",
"description": "Multi-node TP. Splits the model across GPUs spanning multiple nodes. TP is set to total GPU count across all nodes. Head node serves requests, worker nodes run with --headless. Requires InfiniBand interconnect.\n",
"hardware_match": {
"min_gpus": 2,
"multi_node": true
},
"parallel_flag": "--tensor-parallel-size",
"env": {
"GLOO_SOCKET_IFNAME": "$IFACE_NAME",
"NCCL_SOCKET_IFNAME": "$IFACE_NAME"
}
},
"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_tp": "/inclusionAI/Ring-2.6-1T/strategies/single_node_tp.json",
"multi_node_tp_pp": "/inclusionAI/Ring-2.6-1T/strategies/multi_node_tp_pp.json",
"multi_node_tep": "/inclusionAI/Ring-2.6-1T/strategies/multi_node_tep.json",
"multi_node_dep": "/inclusionAI/Ring-2.6-1T/strategies/multi_node_dep.json",
"pd_cluster": "/inclusionAI/Ring-2.6-1T/strategies/pd_cluster.json",
"kv_store_centralized_mooncake": "/inclusionAI/Ring-2.6-1T/strategies/kv_store_centralized_mooncake.json",
"kv_store_distributed_mooncake": "/inclusionAI/Ring-2.6-1T/strategies/kv_store_distributed_mooncake.json"
},
"by_hardware": {
"h200": "/inclusionAI/Ring-2.6-1T/hw/h200.json",
"h100": "/inclusionAI/Ring-2.6-1T/hw/h100.json",
"b200": "/inclusionAI/Ring-2.6-1T/hw/b200.json",
"gb200": "/inclusionAI/Ring-2.6-1T/hw/gb200.json",
"b300": "/inclusionAI/Ring-2.6-1T/hw/b300.json",
"gb300": "/inclusionAI/Ring-2.6-1T/hw/gb300.json",
"mi300x": "/inclusionAI/Ring-2.6-1T/hw/mi300x.json",
"mi325x": "/inclusionAI/Ring-2.6-1T/hw/mi325x.json",
"mi350x": "/inclusionAI/Ring-2.6-1T/hw/mi350x.json",
"mi355x": "/inclusionAI/Ring-2.6-1T/hw/mi355x.json"
}
},
"model": {
"model_id": "inclusionAI/Ring-2.6-1T",
"min_vllm_version": "0.20.2",
"architecture": "moe",
"parameter_count": "1T",
"active_parameters": "50B",
"context_length": 131072,
"base_args": [
"--trust-remote-code",
"--tensor-parallel-size",
"8"
],
"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": "fp8",
"vram_minimum_gb": 1200,
"description": "FP8 weights with TP=8 on B300 or 8x MI300X/MI355X-class nodes"
}
},
"hardware_overrides": {
"amd": {
"extra_env": {
"VLLM_ROCM_USE_AITER": "1"
}
}
},
"guide": "## Overview\n\n[Ring-2.6-1T](https://huggingface.co/inclusionAI/Ring-2.6-1T) is inclusionAI's\nBailingMoeV2_5 FP8 trillion-scale **thinking** model (1T total / 50B active\nparameters) with hybrid linear + MLA attention and a 128K context window.\nIt is the reasoning-focused counterpart to [Ling-2.6-1T](https://huggingface.co/inclusionAI/Ling-2.6-1T)\nin the Ring 2.6 series and emits explicit `...` traces; the\nchat template accepts a `reasoning_effort` field (default `high`).\n\nvLLM 0.20.1 shipped the BailingMoeV2.5 MLA RoPE fix (vllm-project/vllm#41185)\nthat is load-bearing for this architecture — pin v0.20.2 or newer.\n\n## Deployment Configurations\n\n### Docker (AMD MI300X / MI325X / MI355X, TP=8)\n\nTP=8 fits the model-derived 128K context on an MI300X-class node; MI325X\nand MI355X have larger per-GPU HBM and more headroom.\n\n```bash\ndocker run --rm -it \\\n --cap-add=SYS_PTRACE \\\n --ipc=host \\\n --privileged=true \\\n --shm-size=128GB \\\n --network=host \\\n --device=/dev/kfd \\\n --device=/dev/dri \\\n --group-add video \\\n -e VLLM_ROCM_USE_AITER=1 \\\n vllm/vllm-openai-rocm:v0.28.0 \\\n inclusionAI/Ring-2.6-1T \\\n --tensor-parallel-size 8 \\\n --trust-remote-code\n```\n\n### Pip (NVIDIA B300, TP=8)\n\n```bash\nuv venv && source .venv/bin/activate\nuv pip install -U vllm --torch-backend auto\n\nvllm serve inclusionAI/Ring-2.6-1T \\\n --trust-remote-code \\\n --tensor-parallel-size 8\n```\n\n## Client Usage\n\n### Text Generation\n\n```python\nfrom openai import OpenAI\nclient = OpenAI(base_url=\"http://localhost:8000/v1\", api_key=\"EMPTY\")\nresponse = client.chat.completions.create(\n model=\"inclusionAI/Ring-2.6-1T\",\n messages=[{\"role\": \"user\", \"content\": \"Prove there are infinitely many primes.\"}],\n max_tokens=4096,\n temperature=0.6,\n extra_body={\"chat_template_kwargs\": {\"reasoning_effort\": \"high\"}},\n)\nprint(response.choices[0].message.content)\n```\n\nSet `reasoning_effort` to `low` / `medium` / `high` to control the depth of\nthinking. The model wraps its reasoning trace in `...` before\nemitting the final answer.\n\n## References\n\n- [Ring-2.6-1T on Hugging Face](https://huggingface.co/inclusionAI/Ring-2.6-1T)\n- [vLLM PR #41185 — BailingMoeV2.5 MLA RoPE fix](https://github.com/vllm-project/vllm/pull/41185)\n- [vLLM ROCm GPU install docs](https://docs.vllm.ai/en/latest/getting_started/installation/gpu/?device=rocm)\n",
"source": {
"source": "vllm",
"repo": "vllm-project/vllm",
"ref": "main",
"sha": "564ac0de1b228568cafcbfbe8ccaccc82d6023eb",
"path": "recipes/models/inclusionAI/Ring-2.6-1T.yaml",
"edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/inclusionAI/Ring-2.6-1T.yaml",
"overlays": []
}
}