HVOY · API detection guide

Multi-dimensional checksSee how an endpoint behaves and where it may be off. Read the report before you connect it.

No login required · use a test-only API key

A check calls the target endpoint and may spend its available quota.

HVOY / MODEL DATA

Model probe snapshot

Opus 5.5

Ranking snapshot
Probing
94sites
Publicly displayed sites
75sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 71 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

GPT 6.2 Astra

Ranking snapshot
Probing
89sites
Publicly displayed sites
83sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 81 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

GPT 6 Astra

Ranking snapshot
Probing
116sites
Publicly displayed sites
88sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 84 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:58 · Channel data updated: 2026-11-22 09:10

Open the full ranking

GPT 6.2 Sol

Ranking snapshot
Probing
89sites
Publicly displayed sites
84sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 82 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

DeepSeek V4.1 Flash

Ranking snapshot
Probing
56sites
Publicly displayed sites
45sites
Site-equal mean uptime
96%

Meeting the recommendation sample threshold: 41 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:56 · Channel data updated: 2026-11-22 09:09

Open the full ranking

GPT 6.1 Sol

Ranking snapshot
Probing
114sites
Publicly displayed sites
84sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 80 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:58 · Channel data updated: 2026-11-22 09:10

Open the full ranking

GPT 6 Sol

Ranking snapshot
Probing
86sites
Publicly displayed sites
62sites
Site-equal mean uptime
97%

Meeting the recommendation sample threshold: 59 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

Gemini 4 Argon

Ranking snapshot
Probing
31sites
Publicly displayed sites
1sites
Site-equal mean uptime
100%

Meeting the recommendation sample threshold: 1 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:39 · Channel data updated: 2026-11-22 09:00

Open the full ranking

GPT 6.1 Terra

Ranking snapshot
Probing
67sites
Publicly displayed sites
39sites
Site-equal mean uptime
97%

Meeting the recommendation sample threshold: 37 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:09

Open the full ranking

GPT 5.6 Sol

Ranking snapshot
Probing
84sites
Publicly displayed sites
70sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 68 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

GPT Image 2

Ranking snapshot
Probing
30sites
Publicly displayed sites
28sites
Site-equal mean uptime
97%

Meeting the recommendation sample threshold: 25 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:56 · Channel data updated: 2026-11-22 09:09

Open the full ranking

Fable 5.2

Ranking snapshot
Probing
51sites
Publicly displayed sites
39sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 36 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:09

Open the full ranking

Fable 5.1

Ranking snapshot
Probing
68sites
Publicly displayed sites
49sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 46 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

Sonnet 5.5

Ranking snapshot
Probing
82sites
Publicly displayed sites
58sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 55 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:10

Open the full ranking

Seedance 2.5

Ranking snapshot
Listed gateways
8sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: Unknown · Channel data updated: 2026-11-22 09:09

Open the full ranking

Gemini 3.8 Flash

Ranking snapshot
Probing
40sites
Publicly displayed sites
21sites
Site-equal mean uptime
84%

Meeting the recommendation sample threshold: 12 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:59 · Channel data updated: 2026-11-22 09:09

Open the full ranking

Grok 4.7

Ranking snapshot
Probing
47sites
Publicly displayed sites
39sites
Site-equal mean uptime
98%

Meeting the recommendation sample threshold: 38 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:52 · Channel data updated: 2026-11-22 09:09

Open the full ranking

DeepSeek V4 Pro

Ranking snapshot
Probing
34sites
Publicly displayed sites
28sites
Site-equal mean uptime
96%

Meeting the recommendation sample threshold: 27 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:56 · Channel data updated: 2026-11-22 09:09

Open the full ranking

GLM 5.3 Flash

Ranking snapshot
Probing
27sites
Publicly displayed sites
22sites
Site-equal mean uptime
96%

Meeting the recommendation sample threshold: 21 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:52 · Channel data updated: 2026-11-22 09:09

Open the full ranking

GPT 5.6 Terra

Ranking snapshot
Probing
65sites
Publicly displayed sites
49sites
Site-equal mean uptime
97%

Meeting the recommendation sample threshold: 47 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:09

Open the full ranking

Kimi K3

Ranking snapshot
Probing
25sites
Publicly displayed sites
11sites
Site-equal mean uptime
100%

Meeting the recommendation sample threshold: 11 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:56 · Channel data updated: 2026-11-22 09:09

Open the full ranking

GLM 5.3

Ranking snapshot
Probing
31sites
Publicly displayed sites
8sites
Site-equal mean uptime
100%

Meeting the recommendation sample threshold: 8 sites

24-hour statistics snapshot; sites deduplicated by ID. Uptime is sample-weighted within each site, then equally averaged across sites. Monitoring includes non-public sites; meeting the sample threshold does not satisfy every recommendation condition.

Latest recorded probe: 2026-11-22 08:57 · Channel data updated: 2026-11-22 09:09

Open the full ranking

01 / When to check

When is it worth checking first?

01

Before connecting a new gateway

Look at the target endpoint before it carries real traffic, so the choice has one more reference.

See how to start
02

The same model feels different on another route

When answer quality or the experience changes, check whether something deserves a closer look.

How the check works
03

It responds, but you still have questions

A successful call is only the start. More technical signals show how the endpoint actually behaves.

How to read a report

02 / Combined check

Judge the result as a whole, not from one field.

Hvoy AI sends structured probes and scores the endpoint from response protocol, response shape, knowledge answers, identity consistency, reasoning traces, and signature fingerprints.

Cross-checking several technical signals helps you notice a suspicious gateway sooner.

What is observedCombined score · cross-check
  • 01Response protocol
  • 02Response shape
  • 03Knowledge answers
  • 04Identity consistency
  • 05Reasoning traces
  • 06Signature fingerprint
Several signals → cross-check → combined result

03 / Reading a report

Start with these three things.

Homepage example · GPT 6 Astra

This is one GPT 6 Astra result. Other models can return different checks. Use the homepage result for the model you actually run.

Detection Result

Detected modelGPT 6 Astra

0%Below Standard
Detected at
Knowledge Recency
Pass
Variant Fingerprint Match
Pass
Model Family Check
Could not be determined this time; information retrieval was abnormal
Reasoning Reliability
Pass
Schema Constraint Adherence
Pass
Tool Argument Fidelity
Pass
Instruction Hierarchy
Pass
Protocol Field Conformance
Pass
Streaming Response Integrity
Pass
Reasoning Effort Match
Suspected counterfeit: reasoning traits belong to another model
Model Fingerprint Comparison
Consistent with the claimed GPT 6 Astra
  • GPT 6 Astra / GPT 6.1 Sol100%
  • GPT 5.6 Luna<1%
  • GPT 6 Sol<1%
Latency
5.8s
Tokens/s
8.4
Input Tokens
6,898
Output Tokens
193
Cache Read Tokens
13,348
01

First: model and check time

This example is GPT 6 Astra at 2026-11-13 18:27. The checker is who issued the report, not the endpoint that was tested.

Do not mix results from different models or different times.

02

Then: the combined result

This example shows 49% and Below Standard. 49% is not a pass-item ratio or a probability that the model is genuine. Passing most rows does not cancel an abnormal row.

Read the combined result. Do not just count the passes.

03

Last: exceptions and this call

The fingerprint matches here, while other rows are still unresolved or suspicious. A 100% fingerprint is not proof of a unique identity. Latency and token figures describe only this call.

One signal does not replace the combined result.

04 / Next

Check the endpoint yourself

Take the reading, then run a check.

Prepare the endpoint, a test-only API key, and the target model. Configure them on the homepage to start.

Start on the homepageNo login · the call may spend the target endpoint's quota