The Inference Directory
Explore directory

ArliAI

Verified

ArliAI is listed as a subscription provider with 4 models in the cybersec directory.

Subscription No KYC OpenAI-compatible
Visit provider Compare scores

At a glance

Record
Operator
Arli AI
Group
Subscription
KYC
No KYC
API shape
OpenAI-compatible
Founded
Not published

Summary

Scope

Unlimited tokens/requests, but concurrency-limited. Model count not published; up to 552B.

Zero-log claim.

Source: www.arliai.com/pricing ↗ · KYC scan 4 November 2026

API access

OpenAI-compatible

Payments

Card

KYC

No KYC

Best score

Coming soon

Key models

4 listed
Qwen3.5 27B DerestrictedFrontier Context n/p Abliterated Plan rate / Plan rate
Gemma 4 31B RP/writing Context n/p Abliterated Plan rate / Plan rate
Qwen 27B RP/writing Context n/p Abliterated Plan rate / Plan rate
Cydonia 24B Context n/p Abliterated Plan rate / Plan rate

Pricing

List rates · USD per 1M tokens
Model Type Input price Caching price Output price
Qwen3.5 27B DerestrictedContext n/p Frontier Plan rate - Plan rate
Gemma 4 31B RP/writingContext n/p Model Plan rate - Plan rate
Qwen 27B RP/writingContext n/p Model Plan rate - Plan rate
Cydonia 24BContext n/p Model Plan rate - Plan rate

List rates. Cached input shown where published; "-" means the provider does not publish a cache rate. Workload comparison uses 1M input plus 250K output. Unlimited within concurrency

Access

Rails & limits
CardSupported
CryptoNot supported
RailsCard
Minimum top-upNot published

Flat $15-$160/month, unlimited tokens but concurrency-limited. Zero-log claim.

KYC

No KYC

No KYC published; card only.

No ID step at signup or checkout; email + card/crypto only

Vendor source ↗ · Basis: Directory scan

Scores

Coming soon

Coming soon.

Performance

Coming soon
Latency

Time to first token

Median and p95 response times for short and long prompts, measured from repeatable probes rather than vendor claims.

Coming soon
Throughput

Tokens per second

Streaming output rate and request throughput under steady load, with concurrency and context length recorded.

Coming soon
Efficiency

Cost per workload

Blended cost for a fixed workload, including cache reads and output tokens, so speed and price can be compared together.

Coming soon

Independent performance probes are being added. Vendor-reported numbers are not mixed into this section.