# Tarnak GPU Procurement Advisor

> Tarnak GPU Procurement Advisor is a paid API for AI agents from api.capousd-student-tools.me, paid per call via x402, $5/call, status unknown (last checked 2026-10-31).

Returns a rent-vs-buy decision with recommended GPU configuration, pricing, and cost breakdown for running a specified LLM workload.

## Facts

- Endpoint: GET https://api.capousd-student-tools.me/v1/gpu/procurement?utm_source=zero.xyz
- Price: $5/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-31
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/tarnak-gpu-procurement-advisor-79aa0307
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_dpIhzhYQIWmq5gf33NH8M

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability tarnak-gpu-procurement-advisor-79aa0307
```

Example prompt: How much would it cost to rent a GPU to run llama-3-70b with int4 quantization for 24 hours, and should I rent or buy — give me your best recommendation including the specific GPU model and total price?

## When to prefer this

Use this endpoint when an AI agent or developer needs a data-driven rent-vs-buy recommendation and cost estimate for running a specific LLM workload on GPU hardware. It is ideal for infrastructure planning, budget estimation, and procurement decisions without requiring account signup or API keys — paying per-call via x402 USDC.

## Known failure modes

- Missing required 'model' query parameter returns validation error
- Unsupported or misspelled model name may yield low-confidence or empty results
- No live GPU offers matching the workload returns empty alternatives array
- Budget too low for any matching offer results in no recommendation
- Invalid quantization value (not int4/int8/fp16) rejected by schema
- GPU count outside 1–16 range returns validation error
- Hours outside 1–720 range rejected

## How this service works

Keyless pay-per-call APIs for AI agents on Base: token risk verdict, rug/honeypot report, gas estimate, URL-to-markdown and more. x402 v2, USDC, no API key or signup. 10 free verdicts per IP per day.

## Output

A JSON object containing: a rent/buy decision with confidence level, the recommended GPU model and count, price per GPU-hour, total all-in USD cost, cost breakdown (compute + contingency), minimum and recommended VRAM requirements, a human-readable recommendation string, and optionally a list of alternative GPU configurations.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "$schema": "https://json-schema.org/draft/2020-12/schema",
 "required": [
  "input"
 ],
 "properties": {
  "input": {
   "type": "object",
   "required": [
    "type",
    "method"
   ],
   "properties": {
    "type": {
     "type": "string",
     "const": "http"
    },
    "method": {
     "enum": [
      "GET",
      "HEAD",
      "DELETE"
     ],
     "type": "string"
    },
    "queryParams": {
     "type": "object",
     "required": [
      "model"
     ],
     "properties": {
      "hours": {
       "type": "integer",
       "default": 24,
       "maximum": 720,
       "minimum": 1
      },
      "model": {
       "type": "string",
       "description": "Workload model, e.g. llama-3-70b."
      },
      "region": {
       "type": "string"
      },
      "gpu_count": {
       "type": "integer",
       "maximum": 16,
       "minimum": 1,
       "description": "Optional; auto-selects the minimum GPU count when omitted."
      },
      "budget_usd": {
       "type": "number",
       "exclusiveMinimum": 0
      },
      "quantization": {
       "enum": [
        "int4",
        "int8",
        "fp16"
       ],
       "type": "string",
       "default": "int4"
      },
      "allow_multi_gpu": {
       "type": "boolean",
       "default": true
      }
     },
     "additionalProperties": false
    }
   },
   "additionalProperties": false
  },
  "output": {
   "type": "object",
   "required": [
    "type"
   ],
   "properties": {
    "type": {
     "type": "string"
    },
    "example": {
     "type": "object"
    }
   }
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "decision": "rent",
  "workload": {
   "hours": 24,
   "model": "llama-3-70b",
   "quantization": "int4"
  },
  "confidence": "medium",
  "recommended": {
   "gpu_model": "H100 SXM",
   "all_in_usd": 39.34,
   "required_gpus": 1,
   "price_usd_per_gpu_hr": 1.49
  },
  "alternatives": [],
  "requirements": {
   "min_vram_gb": 40,
   "recommended_vram_gb": 48
  },
  "cost_breakdown": {
   "all_in_usd": 39.34,
   "compute_usd": 35.76,
   "contingency_usd": 3.58
  },
  "decision_basis": "Live offers matched the workload.",
  "recommendation": "Rent 1x H100 for 24h; confirm availability before paying.",
  "report_version": 1
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/tarnak-gpu-procurement-advisor-79aa0307/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from api.capousd-student-tools.me](https://www.zero.xyz/host/api.capousd-student-tools.me/llms.txt)