Which uncensored LLM API is best?
There isn’t one best uncensored LLM API for everyone. There is a best one for your budget, your payment method and your feature list. This table is the short version. The rest of the guide shows the numbers behind it.
| Provider | Best for | Payment | API format |
|---|---|---|---|
| Refuseless | Abliterated models, paying by card | Card or crypto | OpenAI and Anthropic |
| Unaligned | Lowest token price, crypto users, no email | Crypto only | OpenAI and Anthropic |
| Uncensored.com | Vision and image generation, card only | Card | Chat completions with X-Api-Key header |
| UncensoredLLM | One key for text, image and video | Card or crypto | OpenAI and Anthropic |
| NanoGPT | Widest model range, tiny first deposit | Card or crypto | OpenAI and Anthropic |
Every provider here is prepaid. You add a balance and each request is deducted from it, so there is no monthly contract to compare. That makes it cheap to try two or three before you settle on one.
What makes a good uncensored chat API?
A good uncensored chat API answers your prompts without refusing, writes well enough for your users, costs what you can afford at your traffic, and fits your stack. In practice that comes down to five checks:
- Model type. Abliterated models have refusals removed from the weights. Uncensored fine-tunes are retrained and can write differently. The guide to abliterated models explains the difference.
- Price on your workload. Input and output are billed separately, so the cheapest provider depends on your ratio of the two.
- Compatibility. An OpenAI-compatible endpoint lets you keep your existing SDK and tools.
- Payment. Card, crypto, or both, plus the minimum deposit.
- Terms and data handling. “Uncensored” describes the model. The provider still sets acceptable-use rules and decides what it logs.
This comparison uses the rates, models and features each provider publishes, as recorded in the directory catalog. NSFW APIs hasn’t run endpoint tests, so nothing below rates refusal behavior or writing quality. That part you have to check yourself, and the last section shows how.
Which models do the uncensored LLM APIs host?
Most uncensored LLM APIs in the directory are built around a few open-weight families: GLM, DeepSeek and Kimi. Each provider sells its own builds, usually a cheap “Flash” model and a larger one.
| Provider | Model | Input | Output | Cached input | Context |
|---|---|---|---|---|---|
| Refuseless | DeepSeek V4.1 Flash 1017 Abliterated | $0.12 | $0.45 | $0.025 | 1M |
| Refuseless | GLM 5.3 Flash Abliterated | $0.12 | $0.50 | $0.025 | 1M |
| Refuseless | GLM 5.3 Abliterated | $1.00 | $2.00 | $0.15 | 1M |
| Unaligned | GLM-5.3 Flash Uncensored | $0.063 | $0.10 | $0.004 | 1M |
| Unaligned | GLM-5.3 Uncensored | $0.15 | $0.24 | $0.017 | 1M |
| Unaligned | GLM-5.3 Abliterated | $2.00 | $3.20 | $0.413 | 1M |
| Uncensored.com | DeepSeek V4.1 Flash Cyber Uncensored | $0.96 | $1.92 | — | — |
| Uncensored.com | GLM-5.3 Flash Cyber Uncensored | $0.96 | $1.92 | — | — |
| Uncensored.com | DeepSeek R1 0528 | $0.50 | $2.18 | — | — |
| Uncensored.com | Kimi K3.1 Pro Cyber Uncensored | $4.80 | $24.00 | — | — |
| UncensoredLLM | GLM-5.3 Flash Uncensored | $0.1053 | $0.3685 | — | — |
| UncensoredLLM | GLM-5.3 Uncensored | $0.282 | $0.846 | — | — |
| UncensoredLLM | DeepSeek V4 Pro Uncensored | $0.97 | $1.94 | — | — |
| UncensoredLLM | GLM-5.3 Abliterated | $2.1825 | $4.365 | — | — |
| UncensoredLLM | Kimi K3.1 Pro Uncensored | $1.20 | $6.00 | — | — |
| NanoGPT | GLM 5.3 Flash Uncensored | $0.20 | $0.80 | — | — |
| NanoGPT | GLM 5.3 Uncensored | $1.25 | $2.25 | — | — |
| NanoGPT | Abliterated Large Flash | $1.00 | $2.00 | — | — |
| NanoGPT | Abliterated Model Large V2 | $3.00 | $5.00 | — | — |
Two things stand out. First, the same base model can carry very different prices. A GLM-5.3 abliterated build costs several times more than an uncensored GLM-5.3 fine-tune at the same provider. If you don’t specifically need abliteration, start with the cheaper build and compare.
Second, matching names don’t mean matching models. “GLM-5.3 Flash Uncensored” at one provider and “GLM 5.3 Flash Uncensored” at another are separate builds, made by different people with different methods. Treat them as different models when you test.
The table shows a selection, not the full catalog. Uncensored.com (119 models), UncensoredLLM (171 models) and NanoGPT (1,271 models) list far more models than fit here, including mainstream and reasoning models.
How much does an uncensored LLM API cost per month?
For a small roleplay app sending 50,000 messages a month, the listed models cost from about $20.40 to $1,800.00 a month. The table below prices every model above on two monthly workloads:
- Roleplay app: 50,000 chat turns, each with 6,000 input tokens (character card, history and new message) and 300 output tokens.
- Story tool: 5,000 long replies, each with 1,500 input tokens and 2,000 output tokens.
| Model | Provider | Roleplay app / month | Story tool / month |
|---|---|---|---|
| GLM-5.3 Flash Uncensored | Unaligned | $20.40 | $1.47 |
| GLM-5.3 Flash Uncensored | UncensoredLLM | $37.12 | $4.47 |
| DeepSeek V4.1 Flash 1017 Abliterated | Refuseless | $42.75 | $5.40 |
| GLM 5.3 Flash Abliterated | Refuseless | $43.50 | $5.90 |
| GLM-5.3 Uncensored | Unaligned | $48.60 | $3.53 |
| GLM 5.3 Flash Uncensored | NanoGPT | $72.00 | $9.50 |
| GLM-5.3 Uncensored | UncensoredLLM | $97.29 | $10.58 |
| DeepSeek R1 0528 | Uncensored.com | $182.70 | $25.55 |
| DeepSeek V4.1 Flash Cyber Uncensored | Uncensored.com | $316.80 | $26.40 |
| GLM-5.3 Flash Cyber Uncensored | Uncensored.com | $316.80 | $26.40 |
| DeepSeek V4 Pro Uncensored | UncensoredLLM | $320.10 | $26.68 |
| GLM 5.3 Abliterated | Refuseless | $330.00 | $27.50 |
| Abliterated Large Flash | NanoGPT | $330.00 | $27.50 |
| GLM 5.3 Uncensored | NanoGPT | $408.75 | $31.88 |
| Kimi K3.1 Pro Uncensored | UncensoredLLM | $450.00 | $69.00 |
| GLM-5.3 Abliterated | Unaligned | $648.00 | $47.00 |
| GLM-5.3 Abliterated | UncensoredLLM | $720.23 | $60.02 |
| Abliterated Model Large V2 | NanoGPT | $975.00 | $72.50 |
| Kimi K3.1 Pro Cyber Uncensored | Uncensored.com | $1,800.00 | $276.00 |
The two workloads rank models differently. A model with a low output price climbs the story ranking even if its input price is average, and a model with expensive output drops. Kimi K3.1 builds are the clearest case: their output price is five times their input price, which hurts a story tool far more than a chat app. Price your own input-to-output ratio rather than borrowing someone else’s ranking.
These figures leave out cached input. Refuseless and Unaligned publish cached rates, which can cut the input bill for long chats sharply. The pricing guide works through caching, deposits and hidden costs step by step.
Can you use an uncensored text API with the OpenAI SDK?
Yes, with 4 of the 5 providers here. They expose OpenAI-compatible endpoints, so you change three values in your code (the base URL, the API key and the model ID) and keep everything else.
| Provider | API format | Abliterated models | Vision | Text-to-image |
|---|---|---|---|---|
| Refuseless | OpenAI and Anthropic | Yes | No | No |
| Unaligned | OpenAI and Anthropic | Yes | No | No |
| Uncensored.com | Chat completions with X-Api-Key header | No | Yes | Yes |
| UncensoredLLM | OpenAI and Anthropic | Yes | No | Yes |
| NanoGPT | OpenAI and Anthropic | Yes | Yes | Yes |
Several also accept the Anthropic message format, which helps if your code or tool was written for Claude. Uncensored.com is the exception: its catalog entry lists “Chat completions with X-Api-Key header”, so the key goes in a custom header instead of the usual bearer token. Most OpenAI clients let you add a custom header, but check that your client supports it before you build around it.
Compatibility only covers the request shape. Features like tool calling, JSON mode and streaming vary by model, so confirm the ones you rely on with a few real requests.
How do you pay for an NSFW LLM API?
You prepay. Each provider takes cards, crypto, or both, and sets its own minimum. This often decides the choice before price does.
| Provider | Payment | Minimum to start | Offer |
|---|---|---|---|
| Refuseless | Card or crypto | $10 first deposit for the bonus | $5 bonus credit |
| Unaligned | Crypto only | $5 in crypto | 5% extra credit |
| Uncensored.com | Card | $10 by card | $5 free credits |
| UncensoredLLM | Card or crypto | $10 by card or crypto | Up to +15% credit |
| NanoGPT | Card or crypto | $0.10 in crypto or $1 by card | 5% off usage |
What the offers are worth in practice:
- Refuseless: $5 bonus credit on a first deposit of at least $10. A $10 deposit buys $15 of usage, a one-time 33% discount on that usage.
- Unaligned: 5% extra credit on the first crypto top-up, capped at $100 of bonus. That is a one-time discount of about 4.8% on that top-up.
- Uncensored.com: $5 of API credits for Uncensored.com Premium users.
- UncensoredLLM: +5% credit on a $200 top-up and +15% on $1,000. That works out to about 4.8% and 13% off the token rate for the usage those top-ups pay for.
- NanoGPT: 5% off eligible usage when you sign up through the referral link. Where it applies, it lowers the token rate itself, for example from $0.20 to $0.19 per 1M input tokens.
Unaligned lets you create an account without an email address. Crypto deposits add a network fee that depends on the coin and network. Offers change, so confirm the current terms on the provider’s site before you deposit.
Who each uncensored LLM API is best for
Refuseless: abliterated models, paying by card
Refuseless is the most focused option. It lists abliterated DeepSeek and GLM builds, publishes cached input rates, and speaks both the OpenAI and Anthropic API formats. It takes cards as well as crypto, and its $5 bonus on a $10 first deposit is the largest relative to the deposit of any offer here. See the Refuseless profile for setup steps and current rates.
Unaligned: lowest token price, crypto users, no email
Unaligned has the lowest published rates in the directory and publishes cached input prices too. You can create an account without an email address. The catch is payment: it takes crypto only, so it doesn’t suit you if you need to pay by card or want a company invoice trail. See the Unaligned profile for setup steps and current rates.
Uncensored.com: vision and image generation, card only
Uncensored.com bundles many uncensored models with vision input and image generation. The example models in our catalog are priced higher than the other providers, but its full catalog starts lower, so check its cheaper models before ruling it out. It takes cards only and uses an X-Api-Key header instead of the standard bearer token. See the Uncensored.com profile for setup steps and current rates.
UncensoredLLM: one key for text, image and video
UncensoredLLM fits an app that needs more than chat. The same prepaid key covers its uncensored LLMs plus image and video models, and it adds an auto router, fallbacks and a Batch API. Its flash model sits second on price for roleplay, and the volume bonuses only kick in at large top-ups. See the UncensoredLLM profile for setup steps and current rates.
NanoGPT: widest model range, tiny first deposit
NanoGPT is a gateway more than a specialist. It carries uncensored and abliterated models next to mainstream ones on one key, so you can route easy requests to a mainstream model and others to an uncensored one. You can start with a very small deposit, and the referral discount lowers the rate itself rather than adding one-time credit. See the NanoGPT profile for setup steps and current rates.
How to choose in an afternoon
You don’t need weeks to decide. With small deposits on two or three providers, one afternoon of testing gets you most of the way.
- Filter by payment. Drop any provider you can’t pay the way you want to.
- Price your workload. Log real token counts, or use the shapes above, and find the two or three cheapest models.
- Add one stronger model. Include a larger model so you can see what the extra money buys.
- Write 20 to 50 test prompts. Use real prompts from your app, including ones a standard model refuses and plenty of ordinary ones.
- Score refusals and quality separately. A model that never refuses but writes flat, repetitive replies isn’t the winner.
- Read the terms. Check the acceptable-use rules and logging policy for your actual use case. Sexual content involving minors and intimate images of real people made without consent are illegal and never allowed.
- Keep a fallback. Providers retire and rename models. A second provider on an OpenAI-compatible endpoint is a one-line switch.
Frequently asked questions
What is the best uncensored LLM API?
It depends on your budget and payment method. On published rates, Unaligned is the cheapest for chat and roleplay. If you need to pay by card, look at Refuseless, Uncensored.com, UncensoredLLM and NanoGPT. If you also need images or video on the same key, UncensoredLLM and NanoGPT cover both.
What is the cheapest uncensored chat API?
Unaligned has the lowest published rates in this directory, with GLM-5.3 Flash Uncensored at $0.063 per 1M input tokens and $0.10 per 1M output tokens. A 50,000-turn roleplay month costs about $20.40 on it before caching.
Which NSFW LLM API accepts credit cards?
Refuseless, Uncensored.com, UncensoredLLM and NanoGPT accept card payments. Unaligned takes crypto only.
Can I use an uncensored LLM API with SillyTavern?
Yes. 4 of the 5 listed providers are OpenAI-compatible, so you can add them in SillyTavern as a custom OpenAI-compatible endpoint with the provider’s base URL, your key and a model ID.
Which uncensored LLM API works without an email signup?
Unaligned lets you create an account without an email address and fund it with crypto.
Does an uncensored LLM API allow any content?
No. Uncensored describes the model’s tendency to refuse, not the provider’s rules. Each provider has acceptable-use terms, and illegal content stays illegal whatever model produces it.