model-memory-fit

Can this AI model run on my graphics card? Estimate model memory requirements and check whether a quantized LLM fits available GPU VRAM before loading or local inference.

Price: $0.05 USDC equivalent
Endpoint: POST /v1/model-memory-fit
Network: Base mainnet · eip155:8453
Asset: USDC
MCP tool: model-memory-fit
External payability: PASS · 2026-10-08T19:55:01.298Z

When an agent should use it

will this LLM fit in my GPU VRAM
check whether a quantized model fits on an 11GB GPU
estimate model memory requirements before loading it
GPU VRAM calculator for local language model inference

Input schema

{
  "type": "object",
  "properties": {
    "parameter_count_b": {
      "type": "number",
      "minimum": 0.01,
      "maximum": 10000
    },
    "quantization_bits": {
      "type": "number",
      "minimum": 1,
      "maximum": 32
    },
    "vram_gb": {
      "type": "number",
      "minimum": 0.1,
      "maximum": 10000
    },
    "runtime_overhead_ratio": {
      "type": "number",
      "minimum": 0,
      "maximum": 2
    }
  },
  "required": [
    "parameter_count_b",
    "quantization_bits",
    "vram_gb"
  ],
  "additionalProperties": false
}

Example request

{
  "parameter_count_b": 7,
  "quantization_bits": 4,
  "vram_gb": 11,
  "runtime_overhead_ratio": 0.15
}

Example output

{
  "verdict": "FIT",
  "estimated_weight_gib": 3.26,
  "estimated_total_gib": 3.75,
  "vram_gib": 11,
  "headroom_gib": 7.25,
  "limitations": [
    "KV cache and architecture-specific buffers are not modeled"
  ]
}

x402 flow

POST /v1/model-memory-fit
→ HTTP 402 + PAYMENT-REQUIRED
→ buyer authorizes locally
→ retry same request with PAYMENT-SIGNATURE
→ one governed execution + machine-readable JSON

Discovery: OpenAPI · x402 · llms.txt

A valid 402/payability probe is operational evidence, not proof of independent demand, future settlement, or semantic correctness.