LLM
Muse Glimmer 30b
Muse Glimmer 30bMuse Glimmer 30B is Meta’s multimodal agent model for multi-step reasoning, reliable tool use, coding, and autonomous tasks.
Model details
View repositoryMuse Glimmer 30B is Meta’s 29.6-billion-parameter multimodal reasoning model for text and image understanding. It supports a 131,072-token context window, configurable reasoning strength, and ATEM tool calling, making it well suited to visual question answering, document analysis, long-context research, and agentic workflows.
This deployment runs the BF16 model on a single H100 through an OpenAI-compatible vLLM API. It pairs Muse Glimmer with its 2.56-billion-parameter DFlash assistant for faster generation while retaining the model’s full context window and multimodal capabilities.
Input
1import os
2from openai import OpenAI
3
4model_id = os.environ["BASETEN_MODEL_ID"]
5
6client = OpenAI(
7 api_key=os.environ["BASETEN_API_KEY"],
8 base_url=(
9 f"https://model-{model_id}.api.baseten.co/"
10 "environments/production/sync/v1"
11 )
12)
13
14response = client.chat.completions.create(
15 model="meta-models/Muse-Glimmer-30B",
16 messages=[
17 {
18 "role": "system",
19 "content": "Reasoning strength: low",
20 },
21 {
22 "role": "user",
23 "content": "Explain quantum computing in simple terms.",
24 },
25 ],
26 max_tokens=512,
27 temperature=1.0,
28 top_p=0.95,
29 extra_body={"top_k": 64},
30)
31
32print(response.choices[0].message.content)
