Arli AI

BILLINGFlat-rate, unlimited tokens/requests on all paid plans; 10% off annual. One-time (non-recurring) payment also offered per Terms. Free $0: max 12K ctx, 1 request at a time, delayed responses, no image gen; docs: free accounts limited to 5 requests per model every 2 days. Starter $10/mo: models up to 31B, 16384 ctx, 1 parallel request, standard priority, image gen (3 MP). Core $15/mo: models up to 355B, 32768 ctx, 1 parallel. Plus $20/mo: up to 355B, 131072 ctx, 1 parallel. Pro $30/mo: up to 355B, 262144 ctx, 2 parallel, high priority, 6 MP images. Max $160/mo: up to 355B, 524288 ctx, 6 parallel, high priority. Starter covers 132 of 137 models (all Gemma-4-31B / Qwen3.5-27B / Qwen3.8-27B variants); Core+ adds the 5 large ones: DeepSeek-V4-Flash-0731 (284B), GLM-4.6-Derestricted-v5 (355B), GLM-4.7 (355B), MiMo-V2.6-Flash-MOPD (310B), Qwen3.8-Flash-Next (180B). API is OpenAI-compatible (https://api.arliai.com/v1/chat/completions, /v1/completions); zero-log. Payment methods: not named publicly — Terms only say "third-party payment processors"; refunds prorated, 30-day satisfaction guarantee.

Per 1K chats = cost of 1,000 messages of ~1,000 tokens in and ~500 out. Click a model to compare it across providers.All models →

Method

Mixed

Access

Web chat + OpenAI-compatible API

Logging

No logs

Trains on prompts

Unverified

Guarantee

Policy only

Jurisdiction

Not disclosed

Signup

Account required

Payments

Unverified

Data checked

Oct 11, 2026

PRIVACYStates it does not store, retain, or have access to prompts or outputs. Usage metadata is kept up to 60 days. All inference runs in-house on its own servers. Training use and jurisdiction aren't stated. Policy ↗

Composite: pending capability and compliance runs · method