TheDrummer: UnslopNemo 12B

TheDrummer: UnslopNemo 12B is a text model from TheDrummer, the creator of Rocinante. It was released by the provider on 2024-11-08, and its provider description frames it as UnslopNemo v4.1, the latest addition to that line, designed for adventure writing and role-play scenarios. It sits in the text category, so it is aimed at language work rather than images, audio or other modalities.

On the input side the model accepts text only, and it produces text only. Its context window is 1,024,000 tokens, which is unusually large and lets you keep long manuscripts, extended chat histories or big reference documents in a single request. The maximum output is 819,200 tokens, so long generations are possible within that window. Its knowledge cutoff is 2024-04-30, which matters if your task depends on events after that date. Supported request parameters include frequency_penalty, logit_bias, logprobs, max_tokens, presence_penalty, repetition_penalty, seed, stop, structured_outputs, temperature, top_k, top_logprobs and top_p.

The stated focus on adventure writing and role-play suggests where this model fits best. Interactive fiction and text adventures are a natural match, since the model can carry a long scenario forward and respond in character. Role-play chat, including persistent characters with extensive backstory, benefits from the large context window because you can keep prior turns and setting notes in the prompt. Longer form creative writing, such as serialized stories or campaign material, is another reasonable use. The sampling parameters give you control over tone and repetition, which helps when a character or narrator voice needs to stay consistent across a long session. Structured outputs and logprobs are available if you want machine readable responses or want to inspect how confident the model was at each step.

You call the model through api14's OpenAI-compatible API using the model id "thedrummer/unslopnemo-12b". Because the interface follows the OpenAI chat completions shape, you can point an existing OpenAI client at the api14 base URL and pass that id in the model field. The request parameters listed above are supported, so you can set temperature, top_p, top_k, repetition_penalty, presence_penalty, frequency_penalty, max_tokens, stop, seed, logit_bias, logprobs, top_logprobs and structured_outputs as needed. The seed parameter is useful when you want reproducible sampling during testing.

Keep the knowledge cutoff in mind: anything after 2024-04-30 is outside what the model learned, so tasks that depend on recent information may need another approach. The model is text only, so if your application needs to handle images or audio, a different model would be a better fit. For text generation, role-play and long context writing work, it is a straightforward option to try.