Agnaistic, usually called Agnai, is an open-source roleplay chat front-end. Besides connecting to other providers, it hosts its own selection of community roleplay fine-tunes. Some are free to use as a guest with no account, and Patreon tiers unlock larger models, longer context, and an OpenAI-compatible API. We list 11 models, all fine-tuned.
This review is based on Agnai’s published model settings, Patreon tiers, and policies, checked on October 11, 2026. Benchmark results from our own runs will be added when they’re complete. Live details are on the Agnai provider page.
The short version
| Best for | Trying roleplay fine-tunes for free with no signup, then paying a few dollars a month for 70B models |
| Uncensored models | 11 (all fine-tuned) |
| Per-token price | None: free models and Patreon tiers |
| Plans | Free (guest models); Patreon from $5/mo up to $50/mo |
| Privacy score | 18 / 100, joint 31st of 37 |
| Watch out for | A generic privacy policy that keeps content and allows research use, plus ad scripts on the site |
Models
Agnai’s hosted models are well-known community roleplay fine-tunes. The ones we list fall into two groups.
Small models (8B to 12B), mostly free:
- Hermes v3 and Stheno v3.4, both on Llama 3.1 8B.
- NemoMix Unleashed, UnslopNemo v4.1 and Lyra v4, all on Mistral Nemo 12B.
- Gemmasutra, on Gemma 9B.
These have an 8,192-token context window in our data. Several of them, including Hermes v3, Stheno v3.4, NemoMix Unleashed and Gemmasutra, are marked for guest use in Agnai’s public settings, so you can chat without an account.
Large models (70B to 72B), paid:
- Hermes 3 (Llama 3.1 70B), with a 16,384-token context, the longest in our list.
- Euryale v2.3, Anubis 70B and EVA v0.1, all on Llama 3.3 70B, at 16,000 tokens.
- Magnum v1 72B (Qwen2 72B), at 12,000 tokens.
Agnai also serves stock (unmodified) models such as Llama, Qwen and DeepSeek. We don’t list those, since this index covers uncensored and roleplay-tuned models. Search the ones we do list on the Models page.
Pricing
There are no per-token prices. Agnai is free to start and funded through Patreon:
- Free: guest models with no account, and a few more if you sign up.
- Patreon, $5 to $50/mo: six tiers at $5, $10, $15, $20, $35 and $50. Paid tiers unlock larger models, longer context, and API access.
The Patreon page doesn’t spell out what each tier includes. Agnai’s public settings show the pattern: the small models are open to everyone, paying members get longer context on them, and the 70B models sit at the highest access levels. Most of the 70B models need more than the $5 tier, so budget for more than the minimum if you want Euryale or Anubis. API access is limited to subscribers.
Because there’s no meter, there’s no cost per 1,000 chats. If you just want to try a few roleplay models, the free guest tier costs nothing and needs no signup.
Privacy
Agnai scores 18 / 100 on our privacy rubric, which only gives points for documented claims:
| Rubric item | Agnai | Points |
|---|---|---|
| Prompt retention | Collects content, logs and device data, kept “as long as needed” | 0 / 20 |
| No training on prompts | Not ruled out: may use data for research | 0 / 15 |
| Technical guarantee (TEE / E2EE) | No | 0 / 20 |
| Signup | None needed for guest models | 8 / 8 |
| Crypto payments | No (card and Patreon) | 0 / 7 |
| Jurisdiction | US (servers) | 10 / 10 |
| Upstream hosts disclosed | No | 0 / 10 |
| Independent audit | No | 0 / 10 |
That’s joint 31st of 37, level with RedModel. The no-signup guest access is a real strength: it earns the full 8 points. The rest of the score reflects the policy. It’s a generic document that says Agnai collects your content, logs and device data, keeps it as long as needed, and may use it for research. The site also runs ad scripts.
We couldn’t find a named legal entity. Contact is through Discord.
Access and payment
- Web chat front-end (open source), with an OpenAI-compatible API for Patreon subscribers.
- No account needed for guest models.
- Paid through Patreon, by card or Patreon’s other methods. No crypto.
- No legal entity named; servers in the US.
How it compares
- Also free, no signup: AI Horde is free and volunteer-run, and we list 17 models there. It scores 28 on privacy, though its own FAQ says to treat requests like public posts.
- Same 70B fine-tunes, clearer terms: Infermatic is $9/mo for API access to Euryale v2.3, Anubis 70B v1.1 and others, with a stated no-logging policy and a privacy score of 40.
- Per token instead: NanoGPT sells Anubis 70B for $0.31 / $0.31 per 1M tokens (input / output), about $0.47 per 1,000 chats, with no account.
Verdict
Agnai is one of the easiest ways to try roleplay fine-tunes. You can start chatting as a guest with no account and no payment, and the Patreon tiers are cheap compared with most flat-rate hosts.
The trade-off is privacy and clarity. The policy keeps your content and allows research use, there’s no named company, and the Patreon page doesn’t say what each tier unlocks. Use it for low-stakes roleplay, and pick a host with a no-logging policy for anything you’d rather keep private.
See Agnai’s models on its provider page, or compare every provider in the index.