positron-ai/Muse-Glimmer-30B

VISIONPricing:Input $1.2 / Cached $0.04 / Output $4.4Concurrent Unit Cost:2Model Size:30BQuant:FP8Context Size:128kPublished:Aug 29, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Muse-Glimmer-30B is a 30 billion parameter language model, originally developed by meta-models, with a notable context length of 131,072 tokens. This model is an unmodified redistribution by Positron AI, maintaining byte-identical weights and configuration from its upstream source. It is primarily intended for use as a pinned, immutable source for build and test machinery, with its original capabilities and evaluations detailed in the upstream model card.

Loading preview...

Overview

Muse-Glimmer-30B is a 30 billion parameter large language model, originally developed by meta-models. This specific repository by Positron AI serves as an unmodified redistribution of the upstream model, ensuring byte-identical weights, tokenizer, and configuration files. It features a substantial context length of 131,072 tokens, indicating its potential for processing extensive inputs.

Key Characteristics

  • Model Origin: Developed by meta-models and redistributed by Positron AI.
  • Parameter Count: 30 billion parameters.
  • Context Length: Supports a very long context window of 131,072 tokens.
  • Immutability: This redistribution is a pinned, immutable snapshot of the original model, ensuring consistency for integration into automated systems.
  • No Modifications: Positron AI has made no modifications to the model's architecture, weights, or training; it is a direct copy of the upstream meta-models/Muse-Glimmer-30B.

Intended Use

This redistribution is primarily intended for use within Positron AI's build and test machinery (CI) as a stable and reliable source. For detailed information regarding the model's specific capabilities, intended applications, and performance evaluations, users are directed to consult the original upstream model card provided by meta-models.