ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Python ai-agentsamdcomfyuidockerllama-cppllmlocal-ain8nnvidiaopen-webuiragself-hostedspeech-to-textstrix-halotext-to-speechworkflow-automation
{
"schema": 1,
"description": "Hardware envelopes for scripts/simulate-model-selection.py and tests/test-model-selection-matrix.py: every tier the installers can see, plus the fleet hosts (fleet_hosts). Pixel-default Linux hosts and macOS select without a tier size ceiling; non-Pixel Linux hosts use the tier map's LLM_MODEL_SIZE_MB.",
"min_context": 65536,
"cpu_container_limit_gib": 6.0,
"envelopes": [
{
"id": "nv-4gb-ram16-pixel",
"label": "NVIDIA 4 GB, RAM 16",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 4096,
"ram_gb": 16,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-4gb-ram16-non-pixel",
"label": "NVIDIA 4 GB, RAM 16",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 4096,
"ram_gb": 16,
"tier": "1",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-6gb-ram16-pixel",
"label": "NVIDIA 6 GB, RAM 16",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 6144,
"ram_gb": 16,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-6gb-ram16-non-pixel",
"label": "NVIDIA 6 GB, RAM 16",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 6144,
"ram_gb": 16,
"tier": "1",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-8gb-ram15-pixel",
"label": "NVIDIA 8 GB, RAM 15",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 8188,
"ram_gb": 15,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": [
"windows-laptop"
]
},
{
"id": "nv-8gb-ram15-non-pixel",
"label": "NVIDIA 8 GB, RAM 15",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 8188,
"ram_gb": 15,
"tier": "1",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-8gb-ram12-pixel",
"label": "NVIDIA 8 GB, RAM 12",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 8188,
"ram_gb": 12,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-8gb-ram12-non-pixel",
"label": "NVIDIA 8 GB, RAM 12",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 8188,
"ram_gb": 12,
"tier": "1",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-8gb-ram32-pixel",
"label": "NVIDIA 8 GB, RAM 32",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 8188,
"ram_gb": 32,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-8gb-ram32-non-pixel",
"label": "NVIDIA 8 GB, RAM 32",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 8188,
"ram_gb": 32,
"tier": "1",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-10gb-ram32-pixel",
"label": "NVIDIA 10 GB, RAM 32",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 10240,
"ram_gb": 32,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-10gb-ram32-non-pixel",
"label": "NVIDIA 10 GB, RAM 32",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 10240,
"ram_gb": 32,
"tier": "1",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-12gb-ram32-pixel",
"label": "NVIDIA 12 GB, RAM 32",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 12288,
"ram_gb": 32,
"tier": "2",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-12gb-ram32-non-pixel",
"label": "NVIDIA 12 GB, RAM 32",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 12288,
"ram_gb": 32,
"tier": "2",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-16gb-ram32-pixel",
"label": "NVIDIA 16 GB, RAM 32",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 16376,
"ram_gb": 32,
"tier": "2",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-16gb-ram32-non-pixel",
"label": "NVIDIA 16 GB, RAM 32",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 16376,
"ram_gb": 32,
"tier": "2",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-20gb-ram64-pixel",
"label": "NVIDIA 20 GB, RAM 64",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 20475,
"ram_gb": 64,
"tier": "3",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-20gb-ram64-non-pixel",
"label": "NVIDIA 20 GB, RAM 64",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 20475,
"ram_gb": 64,
"tier": "3",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-24gb-ram64-pixel",
"label": "NVIDIA 24 GB, RAM 64",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 24564,
"ram_gb": 64,
"tier": "3",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-24gb-ram64-non-pixel",
"label": "NVIDIA 24 GB, RAM 64",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 24564,
"ram_gb": 64,
"tier": "3",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-32gb-ram61-pixel",
"label": "NVIDIA 32 GB, RAM 61",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 32607,
"ram_gb": 61,
"tier": "3",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": [
"tower1",
"tower3"
]
},
{
"id": "nv-32gb-ram61-non-pixel",
"label": "NVIDIA 32 GB, RAM 61",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 32607,
"ram_gb": 61,
"tier": "3",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": [
"tower1",
"tower3"
]
},
{
"id": "nv-40gb-ram128-pixel",
"label": "NVIDIA 40 GB, RAM 128",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 40960,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-40gb-ram128-non-pixel",
"label": "NVIDIA 40 GB, RAM 128",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 40960,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-a100-40gb-ram128-pixel",
"label": "NVIDIA A100 40 GB (40,536 MiB), RAM 128",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 40536,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-a100-40gb-ram128-non-pixel",
"label": "NVIDIA A100 40 GB (40,536 MiB), RAM 128",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 40536,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-48gb-ram128-pixel",
"label": "NVIDIA 48 GB, RAM 128",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 49140,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-48gb-ram128-non-pixel",
"label": "NVIDIA 48 GB, RAM 128",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 49140,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-64gb-ram128-pixel",
"label": "NVIDIA 64 GB, RAM 128",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 65214,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-64gb-ram128-non-pixel",
"label": "NVIDIA 64 GB, RAM 128",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 65214,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-80gb-ram256-pixel",
"label": "NVIDIA 80 GB, RAM 256",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 81559,
"ram_gb": 256,
"tier": "4",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-80gb-ram256-non-pixel",
"label": "NVIDIA 80 GB, RAM 256",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 81559,
"ram_gb": 256,
"tier": "4",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-96gb-ram256-pixel",
"label": "NVIDIA 96 GB, RAM 256",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 97887,
"ram_gb": 256,
"tier": "NV_ULTRA",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": [
"tower2"
]
},
{
"id": "nv-96gb-ram256-non-pixel",
"label": "NVIDIA 96 GB, RAM 256",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 97887,
"ram_gb": 256,
"tier": "NV_ULTRA",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-192gb-ram251-pixel",
"label": "NVIDIA 192 GB, RAM 251",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 195774,
"ram_gb": 251,
"tier": "NV_ULTRA",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": [
"tower2"
]
},
{
"id": "nv-192gb-ram251-non-pixel",
"label": "NVIDIA 192 GB, RAM 251",
"platform": "linux",
"host_kind": "non-pixel",
"backend": "nvidia",
"memory_type": "discrete",
"vram_mb": 195774,
"ram_gb": 251,
"tier": "NV_ULTRA",
"host_arch": "amd64",
"ceiling": "tier-map",
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "apple-8gb",
"label": "Apple 8 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 8,
"tier": "0",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-16gb",
"label": "Apple 16 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 16,
"tier": "1",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": [
"mac-mini"
]
},
{
"id": "apple-24gb",
"label": "Apple 24 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 24,
"tier": "1",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-32gb",
"label": "Apple 32 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 32,
"tier": "2",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-36gb",
"label": "Apple 36 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 36,
"tier": "2",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-48gb",
"label": "Apple 48 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 48,
"tier": "3",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-64gb",
"label": "Apple 64 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 64,
"tier": "4",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-96gb",
"label": "Apple 96 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 96,
"tier": "4",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-128gb",
"label": "Apple 128 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 128,
"tier": "4",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "apple-192gb",
"label": "Apple 192 GB",
"platform": "macos",
"host_kind": "macos",
"backend": "apple",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 192,
"tier": "4",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": false,
"fleet_hosts": []
},
{
"id": "amd-8gb",
"label": "AMD discrete 8 GB",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "discrete",
"vram_mb": 8176,
"ram_gb": 32,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "amd-16gb",
"label": "AMD discrete 16 GB",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "discrete",
"vram_mb": 16368,
"ram_gb": 32,
"tier": "2",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "amd-24gb",
"label": "AMD discrete 24 GB",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "discrete",
"vram_mb": 24560,
"ram_gb": 64,
"tier": "3",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "amd-48gb",
"label": "AMD discrete 48 GB",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "discrete",
"vram_mb": 49136,
"ram_gb": 128,
"tier": "4",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "strix-ram62",
"label": "Strix Halo, RAM 62",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 62,
"tier": "SH_COMPACT",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "strix-ram90",
"label": "Strix Halo, RAM 90",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 90,
"tier": "SH_LARGE",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "strix-ram110",
"label": "Strix Halo, RAM 110",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 110,
"tier": "SH_LARGE",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "strix-ram124",
"label": "Strix Halo, RAM 124",
"platform": "linux",
"host_kind": "pixel",
"backend": "amd",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 124,
"tier": "SH_LARGE",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": [
"strix-halo"
]
},
{
"id": "nv-unified-ram16",
"label": "NVIDIA unified arm64, RAM 16",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 16,
"tier": "2",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-unified-ram32",
"label": "NVIDIA unified arm64, RAM 32",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 32,
"tier": "3",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-unified-ram61",
"label": "NVIDIA unified arm64, RAM 61",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 61,
"tier": "4",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "nv-unified-ram119",
"label": "NVIDIA unified arm64, RAM 119",
"platform": "linux",
"host_kind": "pixel",
"backend": "nvidia",
"memory_type": "unified",
"vram_mb": 0,
"ram_gb": 119,
"tier": "NV_ULTRA",
"host_arch": "arm64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": [
"spark"
]
},
{
"id": "cpu-ram8",
"label": "CPU only, RAM 8",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 8,
"tier": "0",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "cpu-ram12",
"label": "CPU only, RAM 12",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 12,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "cpu-ram16",
"label": "CPU only, RAM 16",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 16,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "cpu-ram20",
"label": "CPU only, RAM 20",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 20,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "cpu-ram24",
"label": "CPU only, RAM 24",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 24,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "cpu-ram32",
"label": "CPU only, RAM 32",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 32,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
},
{
"id": "cpu-ram64",
"label": "CPU only, RAM 64",
"platform": "linux",
"host_kind": "pixel",
"backend": "cpu",
"memory_type": "none",
"vram_mb": 0,
"ram_gb": 64,
"tier": "1",
"host_arch": "amd64",
"ceiling": 0,
"hermes_raise": true,
"fleet_hosts": []
}
],
"reference_architecture": {
"_comment": "Architecture for picks that carry no reviewed layout in the catalog (stale defaults the PR demotes). Only the simulation harness's corrected-estimate column reads this; selection never does. Values are from each model's config.json on Hugging Face (revision noted); size_bytes is the LFS size of the catalog GGUF (its sha256 matches gguf_sha256) at the gguf_source revision.",
"phi4-mini-q4": {
"size_bytes": 2491874272,
"gguf_source": "unsloth/Phi-4-mini-instruct-GGUF@78eb92a46fc37e6b524df991ed9aca9bc6aa7b80",
"block_count": 32,
"attention_head_count_kv": 8,
"attention_key_length": 128,
"attention_value_length": 128,
"recurrent_state_bytes": 0,
"source": "microsoft/Phi-4-mini-instruct@cfbefacb99257ffa30c83adab238a50856ac3083"
},
"phi4-q4": {
"size_bytes": 9053114816,
"gguf_source": "bartowski/phi-4-GGUF@19cd65f97c2f1712a81c506611d3f9c94b16a1e1",
"block_count": 40,
"attention_head_count_kv": 10,
"attention_key_length": 128,
"attention_value_length": 128,
"recurrent_state_bytes": 0,
"source": "microsoft/phi-4@2db69c1c3e91a05d2c64a3185acfbaf36f744e25"
},
"deepseek-r1-7b-q4": {
"size_bytes": 4683073248,
"gguf_source": "unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF@097680e4eed7a83b3df6b0bb5e5134099cadf1b0",
"block_count": 28,
"attention_head_count_kv": 4,
"attention_key_length": 128,
"attention_value_length": 128,
"recurrent_state_bytes": 0,
"source": "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B@916b56a44061fd5cd7d6a8fb632557ed4f724f60"
},
"deepseek-r1-70b-q4": {
"size_bytes": 42520396448,
"gguf_source": "unsloth/DeepSeek-R1-Distill-Llama-70B-GGUF@732dd974083ea5877d7b6d788b36fe7c2e5eab36",
"block_count": 80,
"attention_head_count_kv": 8,
"attention_key_length": 128,
"attention_value_length": 128,
"recurrent_state_bytes": 0,
"source": "deepseek-ai/DeepSeek-R1-Distill-Llama-70B@b1c0b44b4369b597ad119a196caf79a9c40e141e"
},
"qwen3-30b-a3b-q4": {
"size_bytes": 18556686912,
"gguf_source": "unsloth/Qwen3-30B-A3B-GGUF@d5b1d57bd0b504ac62ae6c725904e96ef228dc74",
"block_count": 48,
"attention_head_count_kv": 4,
"attention_key_length": 128,
"attention_value_length": 128,
"recurrent_state_bytes": 0,
"source": "Qwen/Qwen3-30B-A3B@ad44e777bcd18fa416d9da3bd8f70d33ebb85d39"
},
"qwen3.5-35b-a3b-q4": {
"size_bytes": 22016023168,
"gguf_source": "unsloth/Qwen3.5-35B-A3B-GGUF@bc014a17be43adabd7066b7a86075ff935c6a4e2",
"block_count": 40,
"attention_layer_count": 10,
"attention_head_count_kv": 2,
"attention_key_length": 256,
"attention_value_length": 256,
"recurrent_state_bytes": 65863680,
"source": "Qwen/Qwen3.5-35B-A3B@59d61f3ce65a6d9863b86d2e96597125219dc754 (same layout as Qwen3.6-35B-A3B)"
}
}
}