Muse-Glimmer-30B-2026-Knowledge
Copyright 2026 OfflineBase

Built with Muse Spark Open.

This product is a fine-tune of Muse Glimmer 30B (meta-models/Muse-Glimmer-30B,
revision cec94fcfd6512f2cfe708138c32533b346ea4f69), licensed by Meta under the Apache
License, Version 2.0 (see LICENSE).

What we changed:
- We trained LoRA weights (rank 64) on the language model's projection layers. This
  repo holds the weights with the LoRA merged in (BF16). The adapter alone is in
  Muse-Glimmer-30B-2026-Knowledge-LoRA, and the quantized files in Muse-Glimmer-30B-2026-Knowledge-GGUF
  include it.
- We did not change the vision encoder, the tokenizer or the model configuration. The
  tokenizer, processor and config files here are Meta's, unchanged.
- tuned-reasoning-off.jinja is modified from Meta's chat_template.jinja. Unless
  allow_reasoning is set, it opens the reply to the user, with no reasoning channel and
  no tool calls. chat_template.jinja in this repo is Meta's template, unchanged.

Part of the training data was written by Muse Spark Open, so this model is a Derivative
under the Muse Model License (see MUSE_MODEL_LICENSE.md, Exhibit A: Muse Spark Open Usage
Policy). Use of this model must also follow the Muse Glimmer Usage Policy (USAGE_POLICY.md).

Our modifications are licensed under the Apache License, Version 2.0 (see LICENSE).

This is not an official Meta release and is not endorsed by Meta. The names Muse, Muse
Glimmer and Muse Spark appear only to say what this model is built on.