unbottled.ai
Start chatting
Back to Blog
Model UpdateNovember 7, 20265 min read

The newest models are in. Uncensored got serious.

Every major family moved up a generation. A second uncensored provider joined. And for the first time, you can generate uncensored images.

Short version: the catalog is current again. 189 models are live right now - 21 of them uncensored.

You don't have to do anything. Had an older model pinned? Your chats move to its successor on their own.

What "uncensored" means here

Three levels. You choose.

Every model carries a content level. Mainstream models keep their provider's filters. Open models are looser. Uncensored models answer without the lecture.

Mainstream

Provider filters stay on. Safe default for work.

66 live

Open

Fewer refusals, still some guardrails.

102 live

Uncensored

No lecture, no refusal theatre. Your call.

21 live

Nobody picks the level for you. Pick a mainstream model for work, an uncensored one for the question nobody else will answer - same chat, same account.

The new lineup

Grouped by family. Older picks in these families now resolve to the new versions.

Claude

Anthropic's new generation, from deep reasoning to fast replies.
  • Claude Opus 5.5
  • Claude Sonnet 5.5
  • Claude Haiku 5.5
  • Claude Fable 5.1

GPT

OpenAI's GPT-6 family. Every tier has a Pro variant.
  • GPT-6 Astra
  • GPT-6.1 Sol
  • GPT-6.1 Terra
  • GPT-6 Luna

Gemini

Google's flagship plus a fast, cheap workhorse.
  • Gemini 4 Argon
  • Gemini 3.8 Flash

Grok

xAI's latest - rated open, not mainstream.
  • Grok 4.7

Open models

Kimi, DeepSeek, GLM and MiniMax - strong and less filtered.
  • Kimi K3.1 Pro
  • Kimi K3
  • DeepSeek V4.1 Flash
  • GLM-5.3
  • GLM-5.3 Flash
  • MiniMax M3

Uncensored got a real upgrade

More models, two providers, and images.

Venice: the backbone, expanded

Venice runs most of our uncensored catalog. Its flagship moved to Venice Uncensored 1.2, and a whole lineup joined:

  • Venice Role Play Uncensored
  • Gemma 4 26B Uncensored
  • Qwen 3.6 Plus Uncensored
  • GLM 4.7 Flash Heretic
  • GLM 5.3 Abliterated
  • DeepSeek V4.1 Flash Uncensored

NanoGPT: a second route

New provider. It brings GLM 5.3 Uncensored and GLM 5.3 Flash Uncensored with a 1M-token context, and it serves DeepSeek V4.1 Flash Uncensored, GLM 5.3 Abliterated and DeepSeek V4.1 Flash Abliterated next to Venice - plus MiMo V2.6 Flash Abliterated, its own. One provider stumbles - the other picks up.

First uncensored image models

Lustify v8 and Chroma, via Venice. One credit per image. Every image model before them was mainstream or open.

GLM 5.3 Abliterated is the least restricted model we have ever listed. It sits behind the uncensored content level - you only get it if you choose it.

Under the hood

For the people who want to know how the swap actually works.

Old picks follow their successor

Older model ids are now marked legacy: hidden from the picker, still resolvable. Pinned picks and saved skills follow a successor map to the current model - but only if that successor can actually be served for you. If it can't, your original pick stays. No silent swap to an unrelated model.

chatModelSuccessors
Claude Opus 4.6  →  Claude Opus 5.5
GPT-5.5  →  GPT-6.1 Sol
Gemini 3 Flash  →  Gemini 3.8 Flash
Grok 4  →  Grok 4.7
GLM-5.2  →  GLM-5.3

Prices that stay current

New keyless price fetchers for Venice and NanoGPT keep prices, context windows and modalities up to date. Missing price data is exactly why the old Venice model had switched itself off - that can't happen quietly anymore.

Tool calling, honestly

NanoGPT models get native tool calls. Venice Uncensored ignored native tools in our tests, so it keeps a prompt-engineered tool path that does work.

Pick a model. Set your filter.

Every model above is live in the chat. Compare them side by side, or just start talking.

Start a chatBrowse all models
Back to Blog