{ "hf_id": "ricursiveAI/Ripple-65B-Preview", "meta": { "title": "Ripple-65B-Preview", "slug": "ripple-65b-preview", "provider": "Ricursive Intelligence", "description": "Ricursive's Ripple-65B Preview, a dense 65B model for reasoning and agentic coding with hybrid sliding-window and NoPE full attention, tool calling, and a DFlash2 speculative drafter.", "date_added": "2026-11-04", "date_updated": "2026-11-04", "tasks": [ "text" ], "related_recipes": [], "hardware": { "b300": "verified" }, "default_hardware": "b300" }, "recommended_command": { "hardware": "b300", "strategy": "single_node_tp", "variant": "default", "node_count": 1, "deploy_type": "single_node", "env": {}, "docker_image": "vllm/vllm-openai:nightly", "command": "vllm serve ricursiveAI/Ripple-65B-Preview-NVFP4 \\\n --tensor-parallel-size 1 \\\n --reasoning-parser ripple \\\n --enable-auto-tool-choice \\\n --tool-call-parser ripple", "argv": [ "vllm", "serve", "ricursiveAI/Ripple-65B-Preview-NVFP4", "--tensor-parallel-size", "1", "--reasoning-parser", "ripple", "--enable-auto-tool-choice", "--tool-call-parser", "ripple" ], "docker_command": "docker run --gpus all \\\n --privileged --ipc=host -p 8000:8000 \\\n -v ~/.cache/huggingface:/root/.cache/huggingface \\\n vllm/vllm-openai:nightly ricursiveAI/Ripple-65B-Preview-NVFP4 \\\n --tensor-parallel-size 1 \\\n --reasoning-parser ripple \\\n --enable-auto-tool-choice \\\n --tool-call-parser ripple", "docker_argv": [ "docker", "run", "--gpus", "all", "--privileged", "--ipc=host", "-p", "8000:8000", "-v", "~/.cache/huggingface:/root/.cache/huggingface", "vllm/vllm-openai:nightly", "ricursiveAI/Ripple-65B-Preview-NVFP4", "--tensor-parallel-size", "1", "--reasoning-parser", "ripple", "--enable-auto-tool-choice", "--tool-call-parser", "ripple" ], "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": "sm103", "display_name": "B300", "description": "NVIDIA B300 SXM 268 GB HBM3e · 8-GPU HGX B300 node (Blackwell Ultra)", "gpu_count": 8, "vram_gb": 2144, "multi_node": false }, "alternatives": { "kv_store_centralized_mooncake": "/ricursiveAI/Ripple-65B-Preview/strategies/kv_store_centralized_mooncake.json", "kv_store_distributed_mooncake": "/ricursiveAI/Ripple-65B-Preview/strategies/kv_store_distributed_mooncake.json" }, "by_hardware": { "b300": "/ricursiveAI/Ripple-65B-Preview/hw/b300.json", "b200": "/ricursiveAI/Ripple-65B-Preview/hw/b200.json", "gb200": "/ricursiveAI/Ripple-65B-Preview/hw/gb200.json", "gb300": "/ricursiveAI/Ripple-65B-Preview/hw/gb300.json" } }, "model": { "model_id": "ricursiveAI/Ripple-65B-Preview", "min_vllm_version": "0.33.0", "nightly_required": true, "docker_image": "vllm/vllm-openai:nightly", "architecture": "dense", "parameter_count": "65B", "active_parameters": "65B", "context_length": 2097152, "base_args": [], "base_env": {}, "install": { "pip": { "command": "uv venv\nsource .venv/bin/activate\nuv pip install -U vllm --pre \\\n --extra-index-url https://wheels.vllm.ai/nightly/cu130 \\\n --extra-index-url https://download.pytorch.org/whl/cu130 \\\n --index-strategy unsafe-best-match", "note": "vLLM 0.33.0 isn't released yet — nightly required. Nightly wheels are CUDA 13 (cu130) only; the host needs an r580+ NVIDIA driver." }, "docker": { "command": "docker pull vllm/vllm-openai:nightly" } } }, "features": { "reasoning": { "description": "Ripple reasoning parser: splits an optional leading `` block into `reasoning_content`.", "args": [ "--reasoning-parser", "ripple" ] }, "tool_calling": { "description": "Ripple tool-call parser with automatic tool choice.", "args": [ "--enable-auto-tool-choice", "--tool-call-parser", "ripple" ] }, "spec_decoding": { "label": "DFlash2", "description": "DFlash2 speculative decoding with Ricursive's drafter.", "args": [ "--speculative-config", "{\"method\":\"dflash\",\"model\":\"ricursiveAI/Ripple-65B-Preview-DFlash2\",\"num_speculative_tokens\":7}" ] } }, "opt_in_features": [ "spec_decoding" ], "variants": { "default": { "model_id": "ricursiveAI/Ripple-65B-Preview-NVFP4", "precision": "nvfp4", "vram_minimum_gb": 69, "description": "ModelOpt NVFP4 MLPs; attention, embeddings and LM head in BF16.", "json": "/ricursiveAI/Ripple-65B-Preview-NVFP4.json" }, "bf16": { "precision": "bf16", "vram_minimum_gb": 156, "description": "Native BF16 checkpoint." } }, "hardware_overrides": {}, "guide": "## Overview\n\n[Ripple-65B Preview](https://huggingface.co/ricursiveAI/Ripple-65B-Preview)\nis a dense 65B model from Ricursive Intelligence. Its 80 layers alternate four\n1024-token sliding-window layers with one NoPE full-attention layer, and it\nsupports a 2M-token context. The NVFP4 checkpoint fits on one B200/B300.\n\nRipple decides per turn whether to reason, so `reasoning_content` is empty\nwhen it answers directly.\n\n## Client usage\n\n```bash\ncurl http://localhost:8000/v1/chat/completions \\\n -H \"Content-Type: application/json\" \\\n -d '{\n \"model\": \"ricursiveAI/Ripple-65B-Preview-NVFP4\",\n \"messages\": [{\"role\": \"user\", \"content\": \"Write a Python function that reverses a linked list.\"}]\n }'\n```\n\n## References\n\n- [Ripple-65B Preview](https://huggingface.co/ricursiveAI/Ripple-65B-Preview)\n- [Ripple-65B Preview NVFP4](https://huggingface.co/ricursiveAI/Ripple-65B-Preview-NVFP4)\n- [DFlash2 drafter](https://huggingface.co/ricursiveAI/Ripple-65B-Preview-DFlash2)\n- [vLLM PR #65745](https://github.com/vllm-project/vllm/pull/65745)\n", "source": { "source": "vllm", "repo": "vllm-project/vllm", "ref": "main", "sha": "03bd3c68487fe2b957058ab2ce4bc5212a91aa89", "path": "recipes/models/ricursiveAI/Ripple-65B-Preview.yaml", "edit_url": "https://github.com/vllm-project/vllm/edit/main/recipes/models/ricursiveAI/Ripple-65B-Preview.yaml", "overlays": [] } }