--- title: "Best models for JanitorAI, ranked by 7-day use" canonical: https://relayswitch.io/best-models/janitorai generated: 2026-11-08T07:18:38.034Z source: https://api.relayswitch.io/v1/catalog base_url: https://api.relayswitch.io/v1 --- Ranked list · feed read 2026-11-08 06:05 UTC # Best models for JanitorAI: ranked by use, priced per message 15 chat models ordered by calls through the API in the last 7 days, each with the price of one 2K-in / 300-out turn at today's rate, its context window, today's real-model check, and the JanitorAI fields to paste. Every "best model for Janitor AI" list online ranks by traffic share or by opinion and stops before the two numbers you need: what one message costs and what to type. This one ranks the chat models by calls through the RelaySwitch API in the last 7 days, from the public feed, and prices one 2K-in / 300-out turn on each at the live rate. The same key runs all of them; only the model id changes. Uncensored and abliterated builds answer a character card as written, with no jailbreak prompt riding along as paid input. The flagships write the longest, most consistent prose and cost more per message. A build marked _thinks first_ spends part of the reply budget on reasoning, so it needs a larger max tokens in JanitorAI. Don't want to set up an app? The same models run as a browser chat with characters at [Unrestrictable](https://unrestrictable.ai/chat?utm_source=relayswitch.io&utm_medium=site&utm_campaign=best-models), no key to paste. GLM-5.3 UncensoredMost called49,901 calls, $0.00082 a message $0.00032Lowest per messageGLM-5.3 Flash Uncensored Qwen3.6 35B-A3BFirst flagship$0.004 a message 1M tokensLargest contextDeepSeek V4.1 Flash 1017 Uncensored Thinking | # | Model | Calls, 7 days | one 2K-in / 300-out turn | 100 messages | Context | Check | | --- | --- | --- | --- | --- | --- | --- | | 1 | [GLM-5.3 Uncensored](https://relayswitch.io/models/glm-5.3-uncensored)`glm-5.3-uncensored` uncensored build | 49,901 | $0.00082 | $0.0818 | — | Stable | | 2 | [GLM-5.3 Flash Uncensored](https://relayswitch.io/models/glm-5.3-flash-uncensored)`glm-5.3-flash-uncensored` uncensored build | 32,884 | $0.00032 | $0.0321 | — | Stable | | 3 | [DeepSeek V4.1 Flash Abliterated](https://relayswitch.io/models/deepseek-v4.1-flash-abliterated)`deepseek-v4.1-flash-abliterated` uncensored build | 32,737 | $0.0022 | $0.2207 | — | Stable | | 4 | [GLM-5.3 Abliterated](https://relayswitch.io/models/glm-5.3-abliterated)`glm-5.3-abliterated` uncensored build | 18,373 | $0.0059 | $0.585 | — | Stable | | 5 | [GLM-5.3 Flash Abliterated](https://relayswitch.io/models/glm-5.3-flash-abliterated)`glm-5.3-flash-abliterated` uncensored build | 18,225 | $0.0015 | $0.1513 | — | Stable | | 6 | [Qwen3.6 35B-A3B](https://relayswitch.io/models/qwen3.6-35b-a3b)`qwen3.6-35b-a3b` flagship | 14,871 | $0.004 | $0.4044 | — | Stable | | 7 | [DeepSeek V4 Pro Uncensored](https://relayswitch.io/models/deepseek-v4-pro-uncensored)`deepseek-v4-pro-uncensored` uncensored build | 14,090 | $0.0025 | $0.2522 | — | Stable | | 8 | [DeepSeek V4.1 Flash 1017 Uncensored Thinking](https://relayswitch.io/models/deepseek-v4.1-flash-1017-uncensored-thinking)`deepseek-v4.1-flash-1017-uncensored-thinking` uncensored build | 8,165 | $0.0026 | $0.264 | 1M tokens | Stable | | 9 | [DeepSeek V4.1 Flash 1017 Uncensored](https://relayswitch.io/models/deepseek-v4.1-flash-1017-uncensored)`deepseek-v4.1-flash-1017-uncensored` uncensored build | 8,100 | $0.0026 | $0.264 | 1M tokens | Stable | | 10 | [GLM 5.3](https://relayswitch.io/models/glm-5.3)`glm-5.3` flagship | 8,068 | $0.00094 | $0.0935 | 1M tokens | Stable | | 11 | [DeepSeek V4.1 Flash 0910 Uncensored](https://relayswitch.io/models/deepseek-v4.1-flash-0910-uncensored)`deepseek-v4.1-flash-0910-uncensored` uncensored build | 6,662 | $0.0026 | $0.264 | 524K tokens | Stable | | 12 | [Gemini 2.5 Flash](https://relayswitch.io/models/gemini-2.5-flash)`gemini-2.5-flash` flagship | 5,092 | $0.0014 | $0.135 | 1M tokens | Stable | | 13 | [DeepSeek V4.1 Flash 1017](https://relayswitch.io/models/deepseek-v4.1-flash-1017)`deepseek-v4.1-flash-1017` flagship | 4,888 | $0.00086 | $0.086 | 1M tokens | Stable | | 14 | [GLM-5.3 Flash Apex Abliterated](https://relayswitch.io/models/glm-5.3-flash-apex-abliterated)`glm-5.3-flash-apex-abliterated` uncensored build | 4,827 | $0.0026 | $0.256 | 1M tokens | Stable | | 15 | [Qwen3.8 27B Uncensored](https://relayswitch.io/models/qwen3.8-27b-uncensored)`qwen3.8-27b-uncensored` uncensored build | 4,797 | $0.0036 | $0.3625 | 33K tokens | Stable | Top 15 chat models by calls through the API in the last 7 days, feed generated 2026-11-08 06:05 UTC. Prices are USD at the live catalog rate for 2,000 tokens in and 300 out; input $0.282 / output $0.846 per million on the first row. Check is today's real-model canary. ## 01How this list is ranked Rank is calls through the API in the last 7 days, read from `GET https://api.relayswitch.io/v1/stats/top-models` (generated 2026-11-08 06:05 UTC) when this page was built, limited to chat models the catalog lists today. The feed counts every client on the key, from character-chat apps to coding agents; it carries no per-app split, so this is what people run through the same URL you will paste, not a poll of JanitorAI users. Price is one 2K-in / 300-out turn: the card, persona and recent history in, one reply out, at the catalog rate at render. The check column is today's verdict from the [daily real-model check](https://relayswitch.io/proof): every model is sent the same canary prompt once a day and its answer is compared with the last one. Nothing on the page is ranked by taste, and no row is sponsored. ## 02JanitorAI settings for any model on this list Three values, the same for every row; the model id is the only thing that changes. The block uses the first-ranked id: text 1`Proxy URL https://api.relayswitch.io/v1/chat/completions` 2`API key sk-relay-...` 3`Model glm-5.3-uncensored` - **Proxy URL** is the full path, `https://api.relayswitch.io/v1/chat/completions`. Janitor posts to the address as given; a URL that stops at `/v1` gets a 404. - **Model Name** is the id in the table, pasted exactly. Janitor sends it as the `model` field, so a display name or a stray space is a 404 too. - The proxy form has no max tokens field. For a build marked _thinks first_, raise max tokens in Janitor's generation settings to 1,000 or more, or the thinking spends the reply budget and the bubble comes back empty. - Click Add, Save Settings, refresh, and check the configuration reads Active. The first message is the real test; Janitor's own quickstart calls the Check Key button unreliable. The [JanitorAI guide](https://relayswitch.io/guides/janitorai) has the menu path, the fields as the app labels them, and the errors by code. Keys are prepaid from $10 by card or crypto in the [console](https://app.relayswitch.io/console); a key can be narrowed to a model list with a USD budget. ## 03What a long chat costs JanitorAI resends the card, the persona and as much history as the context setting allows with every message, so a reply late in a chat sends far more than 2,000 tokens. The same 300\-token reply on the three most-called models at three history sizes: | Tokens sent per message | GLM-5.3 Uncensored | GLM-5.3 Flash Uncensored | DeepSeek V4.1 Flash Abliterated | | --- | --- | --- | --- | | 4,000 | $0.0014 | $0.00053 | $0.0039 | | 8,000 | $0.0025 | $0.00095 | $0.0073 | | 16,000 | $0.0048 | $0.0018 | $0.0141 | USD per message at the live rates. The input side is what grows; a model that reports cached input bills those tokens at its cached rate. The console's Usage page lists every request with its tokens and charge, and each reply carries a signed receipt you can check on the [verify page](https://relayswitch.io/statements/verify). ## 04Sources Calls: the public feed at `https://api.relayswitch.io/v1/stats/top-models`. Prices, context windows and tags: `https://api.relayswitch.io/v1/catalog`. Verdicts: `https://api.relayswitch.io/v1/proof`. All three are read when the page is built and the page is rebuilt every five minutes. Field labels are quoted from each app's own settings panel in the guides, read 2026-11-02. The same list for [SillyTavern](https://relayswitch.io/best-models/sillytavern) and [Chub AI](https://relayswitch.io/best-models/chub). ## Questions Which model is best for Janitor AI? By use, GLM-5.3 Uncensored (glm-5.3-uncensored): the most-called chat model on this API in the last 7 days, at $0.00082 for one 2K-in / 300-out turn today. By price, GLM-5.3 Flash Uncensored at $0.00032. For prose, the first flagship on the list, Qwen3.6 35B-A3B at $0.004 a message. The table ranks 15 by calls so you can pick by the number that matters to you. How is this Janitor AI list ranked? By calls through the RelaySwitch API in the last 7 days, from the public feed at https://api.relayswitch.io/v1/stats/top-models, read when the page was built. Chat models only, from every client (the feed has no per-app split). Prices are the live catalog rates; the check column is today's real-model canary from /proof. Nothing is ranked by opinion. What does a message cost on Janitor AI with these models? From $0.00032 on GLM-5.3 Flash Uncensored to $0.0059 on the dearest row, for one 2K-in / 300-out turn. A $10 top-up covers about 31,142 such messages on GLM-5.3 Flash Uncensored. A long chat resends its history, so the long-chat table prices the same reply at larger history sizes. Are there free models for Janitor AI here? No free tier on the key: every reply bills its tokens from a prepaid balance that starts at $10, card or crypto, no subscription and no KYC. Before the first top-up the console chat runs 12 free turns a day, enough to try a model before typing it into Janitor AI. The lowest-priced rows cost a fraction of a cent a message. What are the Janitor AI settings for a model on this list? Proxy URL https://api.relayswitch.io/v1/chat/completions, API key sk-relay-..., model id from the table. Every id on the page takes the same three values; only the model changes. The JanitorAI guide on this site has the menu path and the errors by code. Which DeepSeek model is best for Janitor AI? The most-called DeepSeek id here is DeepSeek V4.1 Flash Abliterated (deepseek-v4.1-flash-abliterated) at $0.0022 a message, then DeepSeek V4 Pro Uncensored at $0.0025. Uncensored and abliterated DeepSeek builds answer the card as written; the plain ids keep DeepSeek's own filter. What is the best proxy model for Janitor AI on a budget? GLM-5.3 Flash Uncensored (glm-5.3-flash-uncensored) at $0.00032 for one 2K-in / 300-out turn: $10 is about 31,142 messages. The price grows with the history Janitor resends, so keep the context setting modest on a budget. Related - [JanitorAI setup guide](https://relayswitch.io/guides/janitorai) - [Best models for SillyTavern](https://relayswitch.io/best-models/sillytavern) - [Best models for Chub AI](https://relayswitch.io/best-models/chub) - [Every uncensored build](https://relayswitch.io/uncensored) - [All models by calls](https://relayswitch.io/rankings) - [Models and prices](https://relayswitch.io/models) ## One key, every model on this list Prepaid from $10 by card or crypto. No KYC, no subscription. [JanitorAI guide](https://relayswitch.io/guides/janitorai)[Get a key](https://app.relayswitch.io/console)