Will this Hugging Face model run on my hardware? Fit, best quant, tok/s range, launch command.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Run one Hugging Face model against several devices and rank them (best first). Use for "which of my machines should run X" or "is a 4090 or an M4 Max better for X".
context length in tokens (default 32768).
Hugging Face repo id, e.g. "openai/gpt-oss-120b".
comma-separated devices (preset ids or free text), up to 8.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http modelread https://modelread.com/mcp
Run one Hugging Face model against several devices and rank them (best first). Use for "which of my machines should run X" or "is a 4090 or an M4 Max better for X".
| Argument | Type | Description |
|---|---|---|
| ctx | integer | context length in tokens (default 32768). |
| repo | string | Hugging Face repo id, e.g. "openai/gpt-oss-120b". |
| hardware_list | string | comma-separated devices (preset ids or free text), up to 8. |
{
"type": "object",
"properties": {
"ctx": {
"type": "integer",
"default": 32768,
"description": "context length in tokens (default 32768)."
},
"repo": {
"type": "string",
"default": "openai/gpt-oss-120b",
"description": "Hugging Face repo id, e.g. \"openai/gpt-oss-120b\"."
},
"hardware_list": {
"type": "string",
"default": "rtx5090, m4max-128, dgx-spark, h100-sxm",
"description": "comma-separated devices (preset ids or free text), up to 8."
}
}
}Will this Hugging Face model run on this hardware, how fast, with which quant, and with what command? Checks the repo and its quantized derivatives (GGUF / MLX / AWQ / FP8) from their real file sizes and tensor headers. Returns the best variant and alternatives, each with: verdict (fits | fits-tight | offload | no | unknown), decode and prefill tokens/sec as a calibrated range (lo/mid/hi with basis), longest context that fits, memory placement (VRAM / RAM / SSD) and an exact llama.cpp / vLLM / MLX launch command.
| Argument | Type | Description |
|---|---|---|
| ctx | integer | context length in tokens to fit (default 32768). |
| repo | string | Hugging Face repo id, e.g. "Qwen/Qwen3.6-27B" or "unsloth/gemma-4-26B-A4B-it-GGUF". |
| hardware | string | device preset id ("rtx4090", "m4max-128", "dgx-spark", "8x-h100-sxm", "iphone-17pro-12"), |
| kv_cache_dtype | string | KV cache type: "f16", "q8_0" or "q4_0". |
{
"type": "object",
"properties": {
"ctx": {
"type": "integer",
"default": 32768,
"description": "context length in tokens to fit (default 32768)."
},
"repo": {
"type": "string",
"default": "Qwen/Qwen3.6-27B",
"description": "Hugging Face repo id, e.g. \"Qwen/Qwen3.6-27B\" or \"unsloth/gemma-4-26B-A4B-it-GGUF\"."
},
"hardware": {
"type": "string",
"default": "rtx4090",
"description": "device preset id (\"rtx4090\", \"m4max-128\", \"dgx-spark\", \"8x-h100-sxm\", \"iphone-17pro-12\"),"
},
"kv_cache_dtype": {
"enum": [
"f16",
"q8_0",
"q4_0"
],
"type": "string",
"default": "f16",
"description": "KV cache type: \"f16\", \"q8_0\" or \"q4_0\"."
}
}
}Recommend open models for a task on a device, from a natural-language question. Ranks ~20k indexed Hugging Face models by fit, task benchmark percentile (vendor vs independent kept apart), quant quality and speed. Examples: "best coding model for an RTX 4090 + 64GB RAM at 32k context", "Japanese chat on a Galaxy S25 Ultra", "serve 8 users on 2x RTX 3090".
| Argument | Type | Description |
|---|---|---|
| question | string | the request in plain words; name the device in it, or set default_hardware. |
| default_hardware | string | device used when the question names none (preset id, e.g. "m4max-128"). |
{
"type": "object",
"properties": {
"question": {
"type": "string",
"default": "best coding model for an RTX 4090 with 64GB RAM at 32k context",
"description": "the request in plain words; name the device in it, or set default_hardware."
},
"default_hardware": {
"type": "string",
"description": "device used when the question names none (preset id, e.g. \"m4max-128\")."
}
}
}List device presets: GPUs of every vendor, Apple Silicon, multi-GPU nodes, cloud instances (with $/hour), Jetson, phones, tablets, single-board computers and VPS tiers.
| Argument | Type | Description |
|---|---|---|
| limit | integer | maximum rows to return. |
| filter | string | words matched against id / name / vendor / segment, e.g. "h100", "amd datacenter", "phone samsung". |
{
"type": "object",
"properties": {
"limit": {
"type": "integer",
"default": 40,
"description": "maximum rows to return."
},
"filter": {
"type": "string",
"default": "h100",
"description": "words matched against id / name / vendor / segment, e.g. \"h100\", \"amd datacenter\", \"phone samsung\"."
}
}
}Search the modelread index of Hugging Face models by name and task (no fit estimate; fast).
| Argument | Type | Description |
|---|---|---|
| task | string | chat | coding | vision | ocr | embedding | speech-to-text | tts | image-generation. |
| limit | integer | maximum rows (1-100). |
| query | string | words that must appear in the repo id, e.g. "qwen3.6" or "gemma 4". |
{
"type": "object",
"properties": {
"task": {
"enum": [
"chat",
"coding",
"vision",
"ocr",
"embedding",
"speech-to-text",
"tts",
"image-generation"
],
"type": "string",
"default": "chat",
"description": "chat | coding | vision | ocr | embedding | speech-to-text | tts | image-generation."
},
"limit": {
"type": "integer",
"default": 20,
"description": "maximum rows (1-100)."
},
"query": {
"type": "string",
"default": "gemma 4",
"description": "words that must appear in the repo id, e.g. \"qwen3.6\" or \"gemma 4\"."
}
}
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 18, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.