modelprint_

probes
15 probes· 6 from contributors· 7 groups· every probe is one file, reviewed before it merges

A probe asks a model something a serving stack cannot easily fake, and records the answer as a plain value. Two endpoints that return the same values are running the same thing. Every probe is one small file. Anyone can write one, and an accepted probe is credited here and in the repository.

tokenizer

How many tokens a fixed text costs. Different labs train different tokenizers, and the count is very hard to fake.

english pangram tok-english
prompt_tokens for a fixed 212-char text
prompt_tokens for a fixed English text; tokenizers are unique per lab
read the code v1.0.0
chinese paragraph tok-chinese
CJK segmentation differs most between labs
prompt_tokens for fixed CJK text; segmentation differs most between labs
read the code v1.0.0
code snippet tok-code
indentation + symbol handling
prompt_tokens for fixed source code; indentation and symbol handling vary
read the code v1.0.0
emoji + rare unicode tok-emoji
byte-fallback behaviour
prompt_tokens for emoji and rare codepoints; byte-fallback behaviour differs
read the code v1.0.0
template offset template-offset
hidden system-template tokens
hidden serving-template tokens, from two prompts of known size difference
read the code v1.0.0

errors

The exact wording and shape of validation errors. This prose is written by the lab's own engineers, not by a router.

temperature: 2.0 err-temperature
edge value passes gateways; a lab with a lower ceiling answers in its own words
send temperature 2.0; gateways accept it and forward, labs with a lower ceiling answer with their OWN validation prose (the AllThingsIntel trick)
read the code v1.0.0
max_tokens: 10^9 err-maxtokens
the refusal names the lab's real output limit
ask for a billion output tokens; the refusal names the lab's real limit, and that number is vendor DNA (a gateway forwards this, unlike -1)
read the code v1.0.0
error code family err-code-family
numeric codes are vendor DNA (GLM's 1301 gave Ox away)
the SHAPE of error responses: numeric code families vs string codes vs typed objects
read the code v1.0.0

shape

The vocabulary a model uses to say why it stopped.

finish vocabulary finish-vocab
finish_reason values the endpoint emits
which finish_reason value a forced cut-off produces; vocabularies differ per lab
read the code v1.0.0

network

What the routing layer and the response headers reveal about who served the call.

router regionItIsCuthNotCup net-region
the routing layer names its region and the provider that answered
which edge region and which provider actually served each call, from the router's own opt-in metadata snapshot
read the code v1.0.0
generation recordItIsCuthNotCup net-genrecord
the router's own ledger: provider_name + data_region + native tokens
OpenRouter's /generation endpoint names the serving provider, its data region and the NATIVE token counts for one call
read the code v1.2.0
header dnaItIsCuthNotCup net-headerdna
header presence/format is stack DNA (Bedrock, Vertex, Cloudflare…)
which response headers exist and in what format — serving stacks expose different header families, and CORS hides the rest
read the code v1.0.0

timing

Where the real upstream provider sits, measured through refusal paths rather than through a GPU.

path splitpjperez net-pathsplit
names the upstream provider and the payload it refuses; RTT lives in raw
which upstream provider answers a lane and which payload it refuses, plus a GPU-free router->provider RTT measurement in raw
read the code v3.0.0

capability

Hard ceilings such as the largest prompt a host will accept.

context ceilingItIsCuthNotCup cap-contextceiling
64k, 128k, 200k and 256k+ window classes help date the variant
bisect the accepted prompt-size bucket; window classes help distinguish model variants without probing all the way to 1M
read the code v1.0.0

logits

The shape of the probability numbers, when a host exposes them.

logprob geometryItIsCuthNotCup lp-geometry
δ_norm ≈ 0.32 is universal; the residual spread is a personal signature
the normalized 3rd top-logprob gap on pinned continuations. EVT predicts δ≈0.318 for Gumbel-class models; deviations and gap vectors are model-specific (fallrisk.ai protocol)
read the code v1.0.0
want to write one? read CONTRIBUTING.md · every probe passes a security check and a repeatability check before it merges