{ "schema_version": 2, "tier": "public starter profiles", "as_of": "2026-10-01", "libraos_release": "v0.2.0-rc3", "notice": "Starting points only: settings, licence notes and known problems per hardware and task. Validate on your own documents before relying on a model.", "statuses": { "recommended_starting_point": "Recommended starting point for this hardware and task.", "validated": "Validated for this hardware and task; not the recommended starting point.", "validation_stale": "Was validated on an older LibraOS release; re-validation pending.", "not_yet_validated": "Not yet validated.", "crashed_on_this_hardware": "Crashed this hardware in our testing." }, "profiles": [ { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-qwen38-flash-next-hibrid48-vllm", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "myllmbox/Qwen3.8-Flash-Next-hibrid48", "served_name": "Qwen/Qwen3.8-Flash-Next", "status": "recommended_starting_point", "status_text": "Recommended starting point for this hardware and task.", "serving": { "engine": "vLLM 0.30 (myllmbox image v4, MTP K=5)", "recipe": "myllmbox/qwen38-flash-next-recipe@004d19e", "image": "myllmbox/qwen38-flash-next-vllm:v4@sha256:51629f438f5ba3f7a96db110826c783d69b91a6851c43fb07d447644f157dcc4", "flags": [ { "key": "--kv-cache-memory-bytes", "value": "12000000000" }, { "key": "--max-num-seqs", "value": "8" }, { "key": "--max-model-len", "value": "262144" }, { "key": "--default-chat-template-kwargs", "value": "{\"enable_thinking\": false}" } ] }, "settings": [ { "key": "OPENAI_API_BASE", "value": "http://:8000/v1" }, { "key": "OPENAI_MODEL", "value": "Qwen/Qwen3.8-Flash-Next" }, { "key": "LIBRA_OS_BRAIN_MODEL", "value": "Qwen/Qwen3.8-Flash-Next" }, { "key": "LIBRA_OS_SKILL_MODEL", "value": "Qwen/Qwen3.8-Flash-Next" }, { "key": "LIBRA_OS_LLM_MAX_CONTEXT", "value": "262144" }, { "key": "LIBRA_OS_VECTOR_BACKEND", "value": "postgres" }, { "key": "LIBRA_OS_EMBEDDING_API_BASE", "value": "" }, { "key": "LIBRA_OS_EMBEDDING_MODEL", "value": "Qwen/Qwen3-Embedding-8B" }, { "key": "LIBRA_OS_EMBED_DIM", "value": "4096" }, { "key": "LIBRA_OS_MAX_INFLIGHT_PIPELINES", "value": "2" } ], "licence": { "weights": "Qwen Community License", "flags": [ { "code": "hosted_service_needs_separate_licence", "text": "Offering it as a hosted service needs a separate licence." }, { "code": "closed_binary_in_image", "text": "The serving image contains a closed-source binary." } ] }, "red_flags": [], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "vllm", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-qwen36-35b-a3b-nvfp4-vllm", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "unsloth/Qwen3.6-35B-A3B-NVFP4", "served_name": "unsloth/Qwen3.6-35B-A3B-NVFP4", "status": "validated", "status_text": "Validated for this hardware and task; not the recommended starting point.", "serving": { "engine": "vLLM (mia-vllm-gb10-linear-b12x, MTP 2)", "recipe": "MiaAI-Lab/Unsloth-Qwen3.6-35b-NVFP4-DGX-Spark@79c9e6f", "image": "ghcr.io/miaai-lab/mia-vllm-gb10-linear-b12x@sha256:19627342e1da2607f4db50745dca30e57d7dd0ebff06062f03fd69b43a252931", "flags": [ { "key": "--gpu-memory-utilization", "value": "0.60" }, { "key": "--max-model-len", "value": "262144" }, { "key": "--default-chat-template-kwargs", "value": "{\"enable_thinking\":false,\"preserve_thinking\":true}" } ] }, "settings": [ { "key": "OPENAI_API_BASE", "value": "http://:8000/v1" }, { "key": "OPENAI_MODEL", "value": "unsloth/Qwen3.6-35B-A3B-NVFP4" }, { "key": "LIBRA_OS_BRAIN_MODEL", "value": "unsloth/Qwen3.6-35B-A3B-NVFP4" }, { "key": "LIBRA_OS_SKILL_MODEL", "value": "unsloth/Qwen3.6-35B-A3B-NVFP4" }, { "key": "LIBRA_OS_LLM_MAX_CONTEXT", "value": "262144" }, { "key": "LIBRA_OS_VECTOR_BACKEND", "value": "postgres" }, { "key": "LIBRA_OS_EMBEDDING_API_BASE", "value": "" }, { "key": "LIBRA_OS_EMBEDDING_MODEL", "value": "Qwen/Qwen3-Embedding-8B" }, { "key": "LIBRA_OS_EMBED_DIM", "value": "4096" } ], "licence": { "weights": "Qwen Community License", "flags": [ { "code": "hosted_service_needs_separate_licence", "text": "Offering it as a hosted service needs a separate licence." } ] }, "red_flags": [], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "vllm", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-qwen38-flash-next-mia-nvfp4-vllm", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "Mia-AiLab/Qwen3.8-Flash-Next-NVFP4", "served_name": null, "status": "not_yet_validated", "status_text": "Not yet validated.", "serving": { "engine": "vLLM qwen38-flash-next image, MTP 3", "recipe": "MiaAI-Lab/Qwen3.8-Flash-Next-Single-DGX-Spark@7d0712d", "image": "vllm/vllm-openai:qwen38-flash-next@sha256:fc120ece0a388cc0aa1caad4a9f1cd92113484ab7ec2fd0efadd62585be05bf8", "flags": [] }, "settings": [], "licence": { "weights": "Qwen Community License", "flags": [ { "code": "hosted_service_needs_separate_licence", "text": "Offering it as a hosted service needs a separate licence." }, { "code": "recipe_agpl", "text": "The serving recipe is AGPL-3.0." } ] }, "red_flags": [], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "vllm", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-laguna-s21-nvfp4-vllm", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "poolside/Laguna-S-2.1-NVFP4", "served_name": "poolside/Laguna-S-2.1-NVFP4", "status": "not_yet_validated", "status_text": "Not yet validated.", "serving": { "engine": "vLLM 0.25.1 + DFlash", "recipe": "MiaAI-Lab/Laguna-S-2.1-DGX-Spark-RTX-6000-PRO@750a52e", "image": null, "flags": [ { "key": "--gpu-memory-utilization", "value": "0.65" }, { "key": "--max-model-len", "value": "163840" }, { "key": "--max-num-seqs", "value": "4" }, { "key": "--default-chat-template-kwargs", "value": "{\"enable_thinking\":false}" } ] }, "settings": [], "licence": { "weights": "OpenMDW-1.1", "flags": [ { "code": "recipe_has_no_licence", "text": "The serving recipe repository has no licence file." } ] }, "red_flags": [], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "vllm", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-qwen38-27b-nvfp4-sglang", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "RadixArk/Qwen3.8-27B-NVFP4-BF16-LMHead", "served_name": "Qwen/Qwen3.8-27B", "status": "not_yet_validated", "status_text": "Not yet validated.", "serving": { "engine": "SGLang (MTP)", "recipe": "MiaAI-Lab/Qwen3.8-27B-SGLang-DGX-Spark@9fb18ed", "image": "lmsysorg/sglang:qwen38-27b@sha256:febfb971c7352570fc445c466ebd6ffc9d896024958e544a60f2137fd85856b1", "flags": [] }, "settings": [], "licence": { "weights": "Qwen Community License", "flags": [ { "code": "hosted_service_needs_separate_licence", "text": "Offering it as a hosted service needs a separate licence." } ] }, "red_flags": [], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "sglang", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-qwen38-flash-next-mlx4-tensorfold", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "Vontra/Qwen3.8-Flash-Next-MLX-4bit-MTP", "served_name": "qwen/Qwen3.8-Flash-Next-TF", "status": "not_yet_validated", "status_text": "Not yet validated.", "serving": { "engine": "TensorFold v0.3.6.3 (MLX-format weights, MTP)", "recipe": "MiaAI-Lab/Qwen3.8-Flash-Next-Single-DGX-Spark-TensorFold@a3aa898", "image": "ghcr.io/miaai-lab/qwen3.8-flash-next-single-dgx-spark-tensorfold@sha256:fd30b84dce7a43daaa2be3443e1bd0e7c826923e9cb77a995c4720e76c752c64", "flags": [ { "key": "--parallel", "value": "2" } ] }, "settings": [], "licence": { "weights": "Qwen Community License 1.0", "flags": [ { "code": "hosted_service_needs_separate_licence", "text": "Offering it as a hosted service needs a separate licence." } ] }, "red_flags": [], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "tensorfold", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-nemotron35-lightning-nvfp4-sglang", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4", "served_name": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4", "status": "not_yet_validated", "status_text": "Not yet validated.", "serving": { "engine": "SGLang + DSpark draft", "recipe": "MiaAI-Lab/Nemotron3.5-Lightning-DGX-Spark-RTX-5090-6000-PRO@38c16a4", "image": "lmsysorg/sglang@sha256:a04d9a1a7ffe371b05230aecab001d4ba2bfa0e5c137bc56409ecc4cbc3ac864", "flags": [ { "key": "--mem-fraction-static", "value": "0.78" } ] }, "settings": [], "licence": { "weights": "OpenMDW-1.1", "flags": [] }, "red_flags": [ { "code": "does_not_read_documents", "text": "Answers without reading the document." } ], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "sglang", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-ling30-flash-int4-sglang", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "inclusionAI/Ling-3.0-flash-int4", "served_name": "inclusionAI/Ling-3.0-flash-int4", "status": "crashed_on_this_hardware", "status_text": "Crashed this hardware in our testing.", "serving": null, "settings": [], "licence": { "weights": "MIT", "flags": [ { "code": "component_licence_unclear", "text": "A serving component's licence is unclear." }, { "code": "recipe_has_no_licence", "text": "The serving recipe repository has no licence file." } ] }, "red_flags": [ { "code": "crashed_on_hardware", "text": "Crashed this hardware in our testing." } ], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "sglang", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "dgx-spark-gb10.attachment_qa.gb10-muse-glimmer-30b-nvfp4-vllm", "hardware": "dgx-spark-gb10", "hardware_label": "NVIDIA DGX Spark (GB10)", "deployment": "Local model server: model, LibraOS and database on one box; each profile's requirements list any external service", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "RedHatAI/Muse-Glimmer-30B-NVFP4", "served_name": null, "status": "crashed_on_this_hardware", "status_text": "Crashed this hardware in our testing.", "serving": null, "settings": [], "licence": { "weights": "Apache-2.0 + Meta usage policy", "flags": [ { "code": "restricts_legal_financial_use", "text": "Usage policy restricts legal or financial professional use." } ] }, "red_flags": [ { "code": "crashed_on_hardware", "text": "Crashed this hardware in our testing." }, { "code": "licence_restricts_legal_financial", "text": "Licence restricts legal or financial use." } ], "requirements": { "deployment": "hybrid", "gpu": { "vendor": "nvidia", "families": [ "GB10" ], "min_count": 1, "min_memory_gib_per_gpu": null, "min_memory_gib_total": 115, "unified_memory": true }, "system": { "min_ram_gib": null, "arch": [ "arm64" ], "os": [ "linux" ] }, "runtime": { "engine": "vllm", "min_driver": null }, "external_services": [ "gateway_embeddings" ], "libraos": { "min": "v0.1.22", "max": null } } }, { "profile_id": "gateway.attachment_qa.gateway-qwen36-35b-a3b-262k", "hardware": "gateway", "hardware_label": "Hosted model gateway", "deployment": "Hosted: model calls go to the inference gateway", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "Qwen/Qwen3.6-35B-A3B", "served_name": null, "status": "validation_stale", "status_text": "Was validated on an older LibraOS release; re-validation pending.", "serving": null, "settings": [ { "key": "OPENAI_MODEL", "value": "Qwen/Qwen3.6-35B-A3B" }, { "key": "LIBRA_OS_BRAIN_MODEL", "value": "Qwen/Qwen3.6-35B-A3B" }, { "key": "LIBRA_OS_SKILL_MODEL", "value": "Qwen/Qwen3.6-35B-A3B" }, { "key": "LIBRA_OS_LLM_MAX_CONTEXT", "value": "262144" } ], "licence": null, "red_flags": [], "requirements": { "deployment": "gateway", "gpu": null, "system": { "min_ram_gib": null, "arch": null, "os": null }, "runtime": { "engine": null, "min_driver": null }, "external_services": [ "gateway_llm", "gateway_embeddings" ], "libraos": { "min": null, "max": null } } }, { "profile_id": "gateway.attachment_qa.gateway-glm53-flash-1m-cap64k", "hardware": "gateway", "hardware_label": "Hosted model gateway", "deployment": "Hosted: model calls go to the inference gateway", "task": "attachment_qa", "task_label": "Financial-filing Q&A, document attached", "model": "zai-org/GLM-5.3-Flash", "served_name": null, "status": "not_yet_validated", "status_text": "Not yet validated.", "serving": null, "settings": [ { "key": "OPENAI_MODEL", "value": "zai-org/GLM-5.3-Flash" }, { "key": "LIBRA_OS_BRAIN_MODEL", "value": "zai-org/GLM-5.3-Flash" }, { "key": "LIBRA_OS_SKILL_MODEL", "value": "zai-org/GLM-5.3-Flash" }, { "key": "LIBRA_OS_LLM_MAX_CONTEXT", "value": "1048576" }, { "key": "LIBRA_OS_WORKING_SET_FULL_MAX_TOKENS", "value": "65536" } ], "licence": null, "red_flags": [], "requirements": { "deployment": "gateway", "gpu": null, "system": { "min_ram_gib": null, "arch": null, "os": null }, "runtime": { "engine": null, "min_driver": null }, "external_services": [ "gateway_llm", "gateway_embeddings" ], "libraos": { "min": null, "max": null } } } ] }