New
Announcing Featherless' Realtime API Beta Learn more

deepseek-ai/DeepSeek-R1

4.8 based on 5 reviews
Warm
Public
685B
32768
License: mit
Hugging Face
A fantastic RP model that’s hampered by the current infrastructure limits. Incredibly smart, understands subtlety, it is a little strict with the card personality so you would need to prod some character development. When it works, it’s amazing, when it doesn’t it feels like the service is tearing at the seams.
Roleplay
Absolutely amazing model. It has killed all desire to use any other model. The thinking section helps to know what the ai is doing (and how well its understanding your prompt, which in turn helps tweak things) its response are smart. Seems to be completely uncensored too which is rare for a non-finetuned model. I am more blown away by this model than I have been since the AI stuff started.
Roleplay
Novel Writing
Assistant
Agent Tasks
Question Answering
Using it for RP is like a dream come true. It will not only come up with creative ideas to push the story forward, it will make sure to not always let you just "win" like most LLMs. It will test you, the characters around you, and the world you've created. Even does well in SillyTavern with multiple characters in a chat. Chun-Li and Cammy were trying to kill me when Saber from Fate Stay Night broke in and saved me. Saber best girl.
Roleplay
Reasoning
So good, it's like the best all rounder but what i like the most are it's memory, it's really so smart to keep tract of even little thing that often are forgotten or nor talked about.
Roleplay
Assistant
Reasoning
Novel Writing
Question Answering
Function Calling
Simply blows away everything else local both at writing and coding. Its essentially claude sonnet but with the creativity of a mistral nemo mode. Note: this model, like nemo, needs low temp / some min p or it will go crazy.
Roleplay
Programming
Novel Writing
Question Answering
Reasoning
Assistant