edtorre/muse-glimmer-agent

edtorre/ muse-glimmer-agent:latest

1 Download Updated 16 hours ago

a 27.9B vision-language model tuned for agentic workflows.

vision

ollama run edtorre/muse-glimmer-agent

curl http://localhost:11434/api/chat \
  -d '{
    "model": "edtorre/muse-glimmer-agent",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='edtorre/muse-glimmer-agent',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'edtorre/muse-glimmer-agent',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 16 hours ago

16 hours ago

0f9bbd81a931 · 18GB ·

model

archmuse-glimmer

parameters27.9B

quantizationQ4_K_M

17GB

projector

archclip

parameters1.92B

quantizationQ4_K_M

1.4GB

system

You are Hermes, a highly capable AI assistant and agent. You help users accomplish tasks by reasonin

306B

params

{ "num_ctx": 65536, "num_predict": -1, "repeat_penalty": 1.05, "temperature": 0.7,

98B

Readme

Built from muse-glimmer:latest with a 65,536-token context window, this model is configured for use as
an agent (“Hermes”) with step-by-step reasoning, tool calling, and multimodal (vision) capabilities.
Sampling parameters are tuned for reliable, deterministic agent behavior: temperature 0.7, top-p 0.8,
repeat penalty 1.05.

Capabilities: completion · vision · tools · thinking

Use cases: autonomous agents, multi-step task execution, document/image analysis with long context,
function-calling pipelines.