Skip to content

MODELS_ALIGNMENT.md

Latest commit

 

History

History
66 lines (48 loc) · 2.22 KB

MODELS_ALIGNMENT.md

File metadata and controls

66 lines (48 loc) · 2.22 KB

OllamaMax Models Alignment

This document shows the complete alignment of models across all components of OllamaMax.

Unified Model List (15 Models Total)

Coding & Software Engineering (5 models)

  1. qwen2.5-coder:7b - Python, JS, Bash expert
  2. deepseek-coder:6.7b - Best for coding help
  3. deepseek-r1 - DeepSeek R1 reasoning model
  4. glm-4.6 - GLM 4.6 model
  5. deepseek-v3.1 - DeepSeek v3.1

General Chat & Reasoning (4 models)

  1. llama3.1:8b - Like ChatGPT (DEFAULT MODEL) ✓
  2. qwen3:7b - Qwen3 general purpose
  3. gemma2:9b - Fast & smart
  4. mistral:7b - Lightweight & private

Vision-Language & Multimodal (3 models)

  1. llava:7b - Read images, PDFs
  2. moondream:1.8b - Tiny vision model
  3. qwen3-vl - Qwen3 Vision-Language

Lightweight / Edge (2 models)

  1. phi3:mini - Super fast on CPU
  2. tinyllama:1.1b - Runs anywhere

Embeddings & Search (1 model)

  1. nomic-embed-text - Local RAG / search

Component Alignment Status

1. HTML Dropdown (static/index.html)

✅ Updated - Contains all 15 models with proper tags and installation status indicators

2. Go Backend (main.go)

✅ Updated - AVAILABLE_MODELS array contains all 15 models ✅ Updated - Default model set to llama3.1:8b

3. Shell Script (ollama_install_basic.sh)

✅ Updated - Menu shows all 15 models (options A-O) ✅ Updated - Case statement handles all 15 models ✅ Updated - Proper quantization applied based on hardware detection

Model Naming Conventions

  • All models use the exact Ollama naming convention with tags
  • Examples: llama3.1:8b, qwen2.5-coder:7b, deepseek-coder:6.7b
  • Some models don't have size tags: deepseek-r1, glm-4.6, qwen3-vl
  • Quantization is appended automatically by the shell script based on hardware

Installation Status Indicators

The HTML dropdown shows installation status:

  • ✓ = Installed and ready
  • (Not Installed) = Needs to be pulled from Ollama

The Go backend will automatically offer to pull models when selected if they're not installed.

Default Model

llama3.1:8b is set as the default model across all components as it's:

  • Most likely to be pre-installed
  • Best general-purpose balance
  • Similar to ChatGPT in capabilities