Uncensored models on OpenRouter vs a dedicated uncensored API
OpenRouter gives you one API key for hundreds of models from many providers. It is great for switching between mainstream models. It is not a way to get rid of refusals.
Routing does not remove refusals
A router passes your request to the model you pick. If that model was trained to refuse, it still refuses. To get answers without refusals, the model itself has to be uncensored, either fine-tuned or abliterated.
What to check in an uncensored model
- Size. Many uncensored models are small fine-tunes. They are fine for chat, but weaker at long coding and research tasks.
- Context and output limits. Agents read a lot and write a lot. Short limits end runs early.
- Consistency. On a router, the same model can be served by several providers, so speed, quantization and limits can change between requests unless you pin a provider.
What Silk Compute does differently
We serve only uncensored models, and only large ones: Kimi K3.1, DeepSeek V4.1 Flash and GLM-5.3 Flash, abliterated, with 1M tokens of context each. Every model speaks the OpenAI, Anthropic Messages and OpenAI Responses formats, so Claude Code, Codex and OpenCode work out of the box.
Use both
Keep OpenRouter for mainstream models and add Silk Compute for the work they refuse. Both are OpenAI compatible: change the base URL to https://api.silkcompute.com and the model id, and the rest of your code stays the same. The quickstart takes about two minutes.
