payment-ready 2026-08-05, probe currently failing
Pay-per-job GPU compute API. gpu.probe returns a live nvidia-smi GPU inventory ($0.01); llm.generate runs llama3.1:8b text generation ($0.05). USDC via x402 on Base and Polygon; accepts both standard signed-authorization payments (X-PAYMENT / payment-signature, self-settled) and direct-transfer X-Payment-Tx. Open source: https://github.com/ParabellumDevteam/gpu-forge
Pay from $0.01 to $0.05 per request in USDC on Base / Polygon, settled onchain via the x402 protocol, no signup, no API key needed.
listed 2026-07-12 · no on-chain settlement recorded yet
VERDICT
ranking generation 3mixed 56/100 #10 of 10 in Compute by measured score
What the score read on this service: status offline, 22.6% uptime 30d, 1036ms response p95, x402 compliance 14 of 14, $0.01 price against the category, on-chain traction measured. Each of those is measured, and each is published in full below.
The score comes from our own monitoring, never from reviews, operator claims, or a language model: reliability, x402 compliance, price against the category, a deterministic risk flag, and a small on-chain traction term, weighted by ranking generation 3. It is not placement anyone can buy: the paid verify tier is reported beside it and is not one of its inputs. Every component below stays published and read-only.
It is a score inside Compute, and only there: price and speed are scored against the whole category field, every listing in it that carries no danger flag, the scored service included, which here means 10 listings, this one among them. That set is wider than the 10 the position above is counted against, because a listing we have measured too little of to rank still sits in the field the price and the speed are normalized over. The same service in a different field would read differently. The same engine answers GET /api/v1/best, and asked for a different pool (the whole directory, or another filter) it returns a different number for this service under this same ranking generation. Neither is more correct: they answer different questions, and the methodology states which is which.
Bands on the 0 to 100 measured score: strong at 93 and above, solid from 75, mixed from 55, weak below 55. Price and speed are scored against the whole category field, every listing in it that carries no danger flag, the scored service included, so the score places a service in its own field.
ALTERNATIVES IN COMPUTE
Other Compute services in this directory, ordered by the same measured score. The figures are ours: 30-day uptime from our own probes, 30-day settlement volume read on-chain (a conservative undercount, and null where we cannot measure it, never a zero).
| SERVICE | BAND | SCORE | UPTIME 30D | VOL 30D |
|---|---|---|---|---|
| Cloud World Model | strong | 97 | 100% | --- |
| Coprocesseur déterministe | strong | 97 | 99.9% | --- |
| QuickNode | solid | 89 | 100% | $0 |
| Relaystation | solid | 88 | 100% | --- |
| jsonguard | solid | 86 | 99.7% | --- |
ASSESSMENT
updated 5h agoThe parts the verdict above is computed from, each with its proof. Click any chip. Measured values stay read-only; unknown is honest, and an unknown is never counted as a zero.
reliability 22.6%
- uptime 24h
- 0%
- uptime 7d
- 0%
- uptime 30d
- 22.6%
- uptime 90d
- 72.1%
- response p95
- 1036ms
- avg response
- 485ms
- total checks
- 841
Measured on the unpaid 402 handshake, not the paid call. A service can 402 correctly and still fail after payment.
compliance A (14/14), stale (probe failing)
14 of 14 x402 conformance checks pass. Full checklist below.
price $0.01 to $0.05 (p70 in Compute)
- price (min)
- $0.01
- price (max)
- $0.05
- category percentile (min)
- p70 in Compute
- category percentile (max)
- p85 in Compute
- endpoints / prices
- 2 / 2
- model
- tiered
- stability
- 100%
risk clean
No deterministic risk flag. Risk fires only on an exact blocklist match, or a reserved-brand name with a mismatched verified payTo. Never from low uptime, a high price, or a model guess.
- domain age
- 6894d
- registrar
- MarkMonitor Inc.
- hosting
- custom
- domain created
- 2007-10-09
Identity facts, not a risk score.
traction $0.00 30d · 0 buyers
- volume 30d
- $0.00
- buyers 30d
- 0
- settlements 30d
- 0
- measured since
- --- (start of our harvest window)
- last settlement
- ---
- top buyer share
- -
- trend 7d vs 30d
- -
- networks
- eip155:137, eip155:8453
- volume all-time
- $0
- settlements all-time
- 0
- median settlement 30d
- -
- max settlement 30d
- -
- settled via
- -
Conservative undercount: only USDC settlements via facilitators we measure are counted. A measured floor, not an estimate.
Top buyer share is a concentration signal, not part of the ranking score.
WHAT IT DOES
ai-derivedNo AI synthesis has been produced for this service yet.
ENDPOINTS
| METHOD | PATH | DESCRIPTION | PRICE | NETWORK | ASSET | 402 CHANNEL |
|---|---|---|---|---|---|---|
| GET | /v1/jobs/gpu.probe | $0.01 | Base/Polygon | USDC | header | |
| GET | /v1/jobs/llm.generate | $0.05 | Base/Polygon | USDC | header |
REQUEST / RESPONSE EXAMPLE
An unpaid request to GET /v1/jobs/gpu.probe returns HTTP 402 with the payment terms. Settle onchain via your facilitator, then retry with the X-Payment header.
curl -i 'https://forge.parabellum.tech/v1/jobs/gpu.probe'
// 402 response (captured by monitor) · 2 payloads · click to expand
[
{
"error": "X-Payment-Tx header is required",
"accepts": [
{
"asset": "0x833589fcd6edb6e08f4c7c32d4f71b54bda02913",
"extra": {
"name": "USD Coin",
"chainId": 8453,
"version": "2",
"decimals": 6,
"amountUsd": "0.01",
"settlement": "direct-transfer",
"instructions": "Send the exact USDC amount on base (chainId 8453) to payTo, then retry this request with header X-Payment-Tx: 0x<txhash>. Each tx hash is single-use."
},
"payTo": "0xAA0d72E62E9B458Eb1C4eDfD92336c823C992DeB",
"amount": "10000",
"scheme": "exact",
"network": "eip155:8453",
"outputSchema": {
"input": {
"body": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"type": "http",
"method": "POST",
"bodyType": "json"
},
"output": {
"type": "object",
"properties": {
"gpu": {
"type": "string",
"description": "name, VRAM total/used, utilization, temperature"
}
}
}
},
"maxAmountRequired": "10000",
"maxTimeoutSeconds": 300
},
{
"asset": "0x3c499c542cef5e3811e1192ce70d8cc03d5c3359",
"extra": {
"name": "USD Coin",
"chainId": 137,
"version": "2",
"decimals": 6,
"amountUsd": "0.01",
"settlement": "direct-transfer",
"instructions": "Send the exact USDC amount on polygon (chainId 137) to payTo, then retry this request with header X-Payment-Tx: 0x<txhash>. Each tx hash is single-use."
},
"payTo": "0xAA0d72E62E9B458Eb1C4eDfD92336c823C992DeB",
"amount": "10000",
"scheme": "exact",
"network": "eip155:137",
"outputSchema": {
"input": {
"body": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"type": "http",
"method": "POST",
"bodyType": "json"
},
"output": {
"type": "object",
"properties": {
"gpu": {
"type": "string",
"description": "name, VRAM total/used, utilization, temperature"
}
}
}
},
"maxAmountRequired": "10000",
"maxTimeoutSeconds": 300
}
],
"resource": {
"url": "https://forge.parabellum.tech/v1/jobs/gpu.probe",
"mimeType": "application/json",
"description": "gpu.probe: Returns live nvidia-smi output. Cheap way for agents to verify real GPU capacity before buying bigger jobs."
},
"extensions": {
"bazaar": {
"info": {
"input": {
"body": {},
"type": "http",
"method": "POST",
"bodyType": "json"
},
"output": {
"type": "object",
"example": {
"gpu": "NVIDIA RTX 4000 Ada Generation, 20475 MiB total, 2011 MiB used, 3 % util, 41 C"
}
}
},
"schema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"input"
],
"properties": {
"input": {
"type": "object",
"required": [
"type",
"method",
"bodyType",
"body"
],
"properties": {
"body": {
"type": "object",
"properties": {},
"additionalProperties": false
},
"type": {
"type": "string",
"const": "http"
},
"method": {
"enum": [
"POST"
],
"type": "string"
},
"bodyType": {
"enum": [
"json"
],
"type": "string"
}
}
},
"output": {
"type": "object",
"properties": {
"example": {
"type": "object",
"properties": {
"gpu": {
"type": "string",
"description": "name, VRAM total/used, utilization, temperature"
}
}
}
}
}
}
}
}
},
"x402Version": 2
},
{
"error": "X-Payment-Tx header is required",
"accepts": [
{
"asset": "0x833589fcd6edb6e08f4c7c32d4f71b54bda02913",
"extra": {
"name": "USD Coin",
"chainId": 8453,
"version": "2",
"decimals": 6,
"amountUsd": "0.05",
"settlement": "direct-transfer",
"instructions": "Send the exact USDC amount on base (chainId 8453) to payTo, then retry this request with header X-Payment-Tx: 0x<txhash>. Each tx hash is single-use."
},
"payTo": "0xAA0d72E62E9B458Eb1C4eDfD92336c823C992DeB",
"amount": "50000",
"scheme": "exact",
"network": "eip155:8453",
"outputSchema": {
"input": {
"body": {
"type": "object",
"required": [
"prompt"
],
"properties": {
"model": {
"type": "string",
"description": "Optional Ollama model name, defaults to llama3.1:8b"
},
"prompt": {
"type": "string",
"minLength": 1,
"description": "The prompt to generate from"
},
"max_tokens": {
"type": "number",
"description": "Max tokens to generate, capped at 2048"
}
}
},
"type": "http",
"method": "POST",
"bodyType": "json"
},
"output": {
"type": "object",
"properties": {
"model": {
"type": "string"
},
"response": {
"type": "string"
},
"eval_count": {
"type": "number"
}
}
}
},
"maxAmountRequired": "50000",
"maxTimeoutSeconds": 300
},
{
"asset": "0x3c499c542cef5e3811e1192ce70d8cc03d5c3359",
"extra": {
"name": "USD Coin",
"chainId": 137,
"version": "2",
"decimals": 6,
"amountUsd": "0.05",
"settlement": "direct-transfer",
"instructions": "Send the exact USDC amount on polygon (chainId 137) to payTo, then retry this request with header X-Payment-Tx: 0x<txhash>. Each tx hash is single-use."
},
"payTo": "0xAA0d72E62E9B458Eb1C4eDfD92336c823C992DeB",
"amount": "50000",
"scheme": "exact",
"network": "eip155:137",
"outputSchema": {
"input": {
"body": {
"type": "object",
"required": [
"prompt"
],
"properties": {
"model": {
"type": "string",
"description": "Optional Ollama model name, defaults to llama3.1:8b"
},
"prompt": {
"type": "string",
"minLength": 1,
"description": "The prompt to generate from"
},
"max_tokens": {
"type": "number",
"description": "Max tokens to generate, capped at 2048"
}
}
},
"type": "http",
"method": "POST",
"bodyType": "json"
},
"output": {
"type": "object",
"properties": {
"model": {
"type": "string"
},
"response": {
"type": "string"
},
"eval_count": {
"type": "number"
}
}
}
},
"maxAmountRequired": "50000",
"maxTimeoutSeconds": 300
}
],
"resource": {
"url": "https://forge.parabellum.tech/v1/jobs/llm.generate",
"mimeType": "application/json",
"description": "llm.generate: Text generation on a locally hosted open model via Ollama/vLLM (OpenAI-compatible or Ollama API). Body: { model, prompt, max_tokens? }"
},
"extensions": {
"bazaar": {
"info": {
"input": {
"body": {
"prompt": "Say hello in one short sentence.",
"max_tokens": 64
},
"type": "http",
"method": "POST",
"bodyType": "json"
},
"output": {
"type": "object",
"example": {
"model": "llama3.1:8b",
"response": "Hello there, nice to meet you!",
"eval_count": 9
}
}
},
"schema": {
"type": "object",
"$schema": "https://json-schema.org/draft/2020-12/schema",
"required": [
"input"
],
"properties": {
"input": {
"type": "object",
"required": [
"type",
"method",
"bodyType",
"body"
],
"properties": {
"body": {
"type": "object",
"required": [
"prompt"
],
"properties": {
"model": {
"type": "string",
"description": "Optional Ollama model name, defaults to llama3.1:8b"
},
"prompt": {
"type": "string",
"minLength": 1,
"description": "The prompt to generate from"
},
"max_tokens": {
"type": "number",
"description": "Max tokens to generate, capped at 2048"
}
}
},
"type": {
"type": "string",
"const": "http"
},
"method": {
"enum": [
"POST"
],
"type": "string"
},
"bodyType": {
"enum": [
"json"
],
"type": "string"
}
}
},
"output": {
"type": "object",
"properties": {
"example": {
"type": "object",
"properties": {
"model": {
"type": "string"
},
"response": {
"type": "string"
},
"eval_count": {
"type": "number"
}
}
}
}
}
}
}
}
},
"x402Version": 2
}
] OVER TIME
All charts use the 30d selector; each series spans only the dates it has data for. Every series is also served as JSON at /api/v1/services/gpu-forge/price, /scores, /volume and /buyers. On-chain volume and distinct buyers are measured over the service's settlement address and are a conservative undercount (only settlements that reach a measured facilitator are counted). The on-chain series roll up hourly, so the latest day can be up to about an hour behind; distinct buyers are counted per payout address, so a service that settles to more than one address is an upper bound.
Median across 2 endpoints. Use the selector to isolate one.
checklist grew 11->14 on 2026-07-28; a step here is a metric change, not a regression
Measured site and economics pillars from the assessment history, so the latest value shown elsewhere on this page reads as a point on a trend rather than a permanent state.
COMPLIANCE
14/14 checks pass · grade A · stale (probe failing)last 402 captured 2026-07-13 · last up 2026-08-05
- 402 payload captured
- accepts[] array present
- payTo address recoverable
- payTo at accepts[0].payTo (conformant shape)
- payTo is a valid on-chain address
- atomic price declared
- atomic price in a sane range
- asset (token) address declared
- network resolves to CAIP-2
- payment scheme declared
- served over HTTPS
- declares the current x402 version (2)
- EIP-712 domain parameters present on every EVM entry
- x402 v2 envelope delivered in the payment-required header
SITE PILLARS
- homepage reachable
- openapi doc
- pricing page
- llms.txt
- robots.txt
- terms page
recent checks (18) live · click to expand
EMBED THIS BADGE
<a href="https://x402-list.com/services/gpu-forge?utm_source=badge&utm_medium=referral&utm_campaign=embed"> <img src="https://x402-list.com/badge/gpu-forge.svg" alt="GPU Forge listed on x402-list" height="28"> </a>
[](https://x402-list.com/services/gpu-forge?utm_source=badge&utm_medium=referral&utm_campaign=embed)
<a href="https://x402-list.com/services/gpu-forge?utm_source=badge&utm_medium=referral&utm_campaign=embed"> <img src="https://x402-list.com/badge/gpu-forge.svg?data=uptime" alt="GPU Forge uptime on x402-list" height="28"> </a>