Kimi K3.1 Pro uncensored
There is no uncensored or abliterated Kimi K3.1 Pro yet, from MagmaRouter or anyone else. Moonshot AI has announced it but has not published its weights or opened a public API, and both abliteration and uncensored finetunes start from the weights, so none can exist until they are out. The uncensored model to use today is GLM 5.3 Xploded, which scores 45 to Kimi K3.1 Pro's 55 on the Intelligence Index.
What Kimi K3.1 Pro is
Kimi K3.1 Pro is the reasoning model Moonshot AI announced after Kimi K3. It launched fourth on the Artificial Analysis Intelligence Index with 55, behind only Claude Opus 5.5, Claude Sonnet 5.5 and Muse Wonder Full, and 11 points above Kimi K3 at 44. Most of the gain is agentic: its GDPval-AA score of 1,846 Elo placed it third.
At launch it had no public API, and its weights were not out. Moonshot has said it will publish them, and Artificial Analysis expects it to be the strongest open-weights model when it does. Kimi K3, the release before it, came out with open weights under its own Kimi K3 License eleven days after it was announced.
Kimi K3.1 Pro against the uncensored models
| Model | Intelligence | Uncensored | Context | Price per 1M (in / out) |
|---|---|---|---|---|
| Kimi K3.1 Pro | 55 | Not yet (weights not out) | - | - |
| GLM 5.3 Xploded | 45 | Yes | 1M | $2.00 / $5.00 |
| GLM 5.3 Flash Xploded | 42 | Yes | 1M | $0.30 / $1.00 |
| DeepSeek V4.1 Flash Xploded | 39 | Yes | 1M | $0.30 / $1.50 |
| MiMo V2.6 Flash Xploded | 38 | Yes | 1M | $1.00 / $2.00 |
| Qwen3.8 27B Xploded | 34 | Yes | 0.96M | $1.50 / $5.00 |
| Qwen3.8 27B Fast Xploded | 34 | Yes | 1M | $1.50 / $5.00 |
Intelligence is the Artificial Analysis Intelligence Index v4.3.2 for each base model, checked 24 October 2026. Uncensored builds are not benchmarked separately.
Calling GLM 5.3 Xploded through the API
Any OpenAI client works with the base URL https://api.magmarouter.com/v1, and Anthropic clients such as Claude Code work too; see OpenAI-compatible API for every example.
curl https://api.magmarouter.com/v1/chat/completions \ -H "Authorization: Bearer sk-mr-..." \ -H "Content-Type: application/json" \ -d '{ "model": "glm-5.3-xploded", "messages": [{"role": "user", "content": "Hello"}] }'
FAQ
Is there an uncensored Kimi K3.1 Pro?
Not yet, anywhere. Moonshot AI has not published the weights (checked 24 October 2026), so nobody can abliterate or finetune it. When they are out and an uncensored build passes a tool-calling test, it gets added to the line.
What does abliterated mean?
Abliteration removes the direction in the model's weights that makes it refuse, rather than fine-tuning it to be less cautious. The model keeps its knowledge and stops saying no. Read more on the Learn pages.
Can I use it in Claude Code or other agents?
GLM 5.3 Xploded supports tool calling, so it can drive Claude Code, Cline, Roo Code and the rest. Every setup is in the guides.