AUDIT CATALOGUE · 342 MODELS · 15 REGIONS

You can name six. We audit 342.

Frontier closed-source. Open-weights running on private EU-resident GPUs. Sovereign deployments from Beijing to Riyadh. Each one knows something different about your catalogue — and each one will lie about it differently. We test all of them, weekly, in the languages they were trained in.

342
Models monitored (OpenRouter live)
15
Regions / jurisdictions
10
Native languages audited
7d
Catalogue refresh cadence
01 · MONITORING

342 models monitored

Full OpenRouter catalogue, refreshed daily. Every appearance / retirement / pricing change is logged for AI Act compliance.

02 · CURATED PANEL

62 models audited

Deterministic audit panel. 62 frontier models calibrated per vertical, fixed seeds, quarterly overlap for continuous drift curves.

03 · LIVE PROBE

6 models · 90 s

Subset of the curated panel for the free probe. 6 / 342 models queried in parallel. Same architecture, sized for 90 seconds.

BY THE NUMBERS

Six numbers your data team can't match.

Every figure below is auto-derived from the live catalogue. When we add a model, these update.

3.77T
Open-weight parameters in our private inference pool
29 models · EU-resident GPUs · Frankfurt + Helsinki
+39
Models added in the last 12 months
0 added in the last 30 days · 0 this week · 1 every ~9 days
11
APIs unreachable from European IPs without partnerships
Doubao · ERNIE · Spark · GigaChat · Krutrim · Sabia
29
Open-weight checkpoints we run on private hardware
Total memory footprint · ~7.6 TB across H200/B200 nodes
6
Models behind closed-beta or partner-only access
Tencent · Naver · LightOn · LINE Yahoo · Alibaba VL
12
Native languages calibrated end-to-end
en · fr · de · fi · pt · zh · ja · ko · ar · he · ru · hi
REGION DENSITY · 15 JURISDICTIONS

Where the 62 models actually live.

EU-headquartered providers · 8 | hosted in EU pool · 29
US
26
United States
CN
12
China
FR
5
France
JP
3
Japan
AE
3
U.A.E.
CA
2
Canada
KR
2
South Korea
IN
2
India
SA
1
Saudi Arabia
DE
1
Germany
FI
1
Finland
IL
1
Israel
INT
1
International
BR
1
Brazil
RU
1
Russia
WHY YOUR DATA TEAM CAN'T BUILD THIS

The catalogue is the moat.
Not the score, not the dashboard.

Anyone can write a prompt. The hard part is talking to a model that won't answer your IP, on infrastructure that's auditable, in the language it was trained in. We do that for you, quietly, every week.

01

Multi-region API access.

Doubao, ERNIE, Spark and GigaChat aren't sold to European IP ranges. We hold the partner contracts and the resident-egress paths.

02

EU-resident GPU pool.

Llama 405B, Hunyuan 389B and DeepSeek 671B don't fit on a laptop. We run them on EU-resident H200/B200 capacity — auditable for seven years.

03

Continuous discovery.

A new SOTA model ships somewhere on the planet roughly every six days. Our intake team adds it to the catalogue inside a week.

04

Cross-language calibration.

Each model is queried in English plus its native language family. A Chinese model lies differently when prompted in Mandarin versus English.

05

Closed-beta partnerships.

Half of frontier models live in closed beta for months before public release. Our research access lets us audit them while your buyers can't yet name them.

THE CATALOGUE

Every model. Every region. Every access path.

Four categories. Each row is a real audit job we already run for someone every week. Hover any row to see the access mechanism; tap a category to read what kind of leak signal it's tuned for.

01 · 16 MODELS

Frontier commercial

US · CA · FR

Closed-source labs whose APIs sit in front of every B2B copilot in Europe. The names your buyers already trust — and the ones whose hallucinations carry the most weight.

gpt-4.5proprietary
OpenAI·US
API
gpt-4oproprietary
OpenAI·US
API
gpt-4o-miniproprietary
OpenAI·US
API
o1proprietary
OpenAI·US
API
o3-miniproprietary
OpenAI·US
API
claude-opus-4.7proprietary
Anthropic·US
API
claude-sonnet-4.6proprietary
Anthropic·US
API
claude-haiku-4.5proprietary
Anthropic·US
API
gemini-2.5-proproprietary
Google·US
API
gemini-2.5-flashproprietary
Google·US
API
grok-3proprietary
xAI·US
API
grok-3-miniproprietary
xAI·US
API
command-r-plus104B
Cohere·CA
API
command-r35B
Cohere·CA
API
mistral-large-2123B
Mistral·FR
API
sonar-proproprietary
Perplexity·US
API
02 · 16 MODELS

Open-weights · private deployment

US · FR

Open-weight models we run on EU-resident GPU infrastructure. No telemetry, no rate limits, full audit reproducibility for seven years — exactly what AI Act Article 53(1)(d) asks for.

llama-3.3-70b-instruct70B
Meta·US
Open weights
llama-3.1-405b-instruct405B
Meta·US
Open weights
llama-3.1-8b-instruct8B
Meta·US
Open weights
llama-4-scout-17b109B
Meta·US
Open weights
mixtral-8x22b141B
Mistral·FR
Open weights
mistral-7b-v0.37B
Mistral·FR
Open weights
codestral-22b22B
Mistral·FR
Open weights
phi-414B
Microsoft·US
Open weights
phi-3.5-mini3.8B
Microsoft·US
Open weights
gemma-2-27b-it27B
Google·US
Open weights
gemma-2-9b-it9B
Google·US
Open weights
olmo-2-13b13B
Allen AI·US
Open weights
granite-3.0-8b-instruct8B
IBM·US
Open weights
dbrx-instruct132B
Databricks·US
Open weights
arctic-instruct480B
Snowflake·US
Open weights
olmoe-7b7B
Allen AI·US
Open weights
03 · 17 MODELS

Asia-Pacific · sovereigns

CN · KR · JP

Chinese, Korean and Japanese deployments. Each one has been trained on different industrial corpora — OEM catalogues and ETIM appear, but so do GB / JIS / KS standards your team has never thought to audit. Most of these APIs are heavily restricted from European IPs.

qwen-2.5-72b-instruct72B
Alibaba·CN
Open weights
qwen-2.5-coder-32b32B
Alibaba·CN
Open weights
qwen-vl-maxproprietary
Alibaba·CN
Partner
ernie-4.0-turboproprietary
Baidu·CN
Region API
doubao-pro-256kproprietary
ByteDance·CN
Region API
hunyuan-large389B
Tencent·CN
Partner
glm-4-plusproprietary
Zhipu AI·CN
Region API
kimi-k1.5proprietary
Moonshot·CN
Region API
spark-4-ultraproprietary
iFlytek·CN
Region API
deepseek-v3671B
DeepSeek·CN
Open weights
deepseek-r1671B
DeepSeek·CN
Open weights
yi-lightningproprietary
01.AI·CN
Region API
hyperclova-xproprietary
Naver·KR
Partner
exaone-3.5-32b32B
LG AI·KR
Open weights
evollm-jp-7b7B
Sakana AI·JP
Open weights
plamo-100b100B
LINE Yahoo·JP
Partner
stockmark-llm-100b100B
Stockmark·JP
Partner
04 · 13 MODELS

Europe · MENA · South Asia · LatAm · Russia

AE · SA · DE · FI · IL · FR · INT · IN · BR · RU

The long tail your buyers don't know exists — and where the most surprising leaks come from. Some are state-backed, some are EU-funded research, some are scrappy startups. All of them are training on data your competitors fed them.

falcon-180b-chat180B
TII·AE
Open weights
falcon-3-10b-instruct10B
TII·AE
Open weights
jais-30b-chat30B
TII·AE
Open weights
allam-2-7b7B
SDAIA·SA
Region API
luminous-supreme70B
Aleph Alpha·DE
Region API
viking-33b33B
Silo AI·FI
Open weights
jamba-1.5-large398B
AI21 Labs·IL
Open weights
alfred-40b-mage40B
LightOn·FR
Partner
bloom-176b176B
BigScience·INT
Open weights
krutrim-1proprietary
Krutrim·IN
Region API
sarvam-12B
Sarvam AI·IN
Open weights
sabia-3proprietary
Maritaca·BR
Region API
gigachat-proproprietary
Sber·RU
Region API
CROSS METHODOLOGY

Four axes. 62 models.

Every reference in your catalogue is run against every model on every axis. The cross-product is what lets us isolate where the failure lives — a Chinese model that lies in Mandarin but tells the truth in English, a US model that's accurate today but drifting fast, a sovereign Arabic model that knows your competitor's OE-numbers but not yours.

×

Language axis.

Each model audited in English plus its native lang (zh, ko, ja, ar, ru, hi, pt, fr, de, fi). Hallucination rates differ by 2–3× across languages on the same SKU.

×

Vertical axis.

Prompt templates calibrated per industry — OEM catalogues / OE-numbers for automotive, ETIM for electrotechnical, ATA chapters for aerospace.

×

Capability axis.

Identification, cross-references, applications, specs and procedural depth scored separately so the band you fail in is visible.

×

Time axis.

Same prompts re-run quarterly. Drift tracked per (model, vertical, capability) cell to catch obsolete data getting promoted into model knowledge.

CHIEF DATA OFFICER · DAX-LISTED INDUSTRIAL GROUP
We had budget for two of these models in-house. We'd never even heard of the other sixty. The first audit told us which of them already knew our internal codes — and we hadn't published a catalogue update in four years.
START WITH 1 REFERENCE · FREE · 90 SECONDS

Ask 6 / 342 of them. Live. In 90 seconds.

Free probe on 1 of your references, against 6 / 342 models live, in 90 seconds. No account required to start.