Skip to content
Justin's Digital Garden

Artificial Intelligence (AI) Models

Model and Use Cases

Section titled “Model and Use Cases”
Model Name Use Cases Strengths Notes
Codebooga Programming Python and JavaScript 1
CodeGeeX Programming Cross language translation, plugins for many IDEs 2
Codeqwen Programming 3
Deepseek-R1 General Reasoning
Dolphin-mistral General Uncensored, use when other models refuse answers 4
DolphinMixtral Programming Uncensored 5
Gemma2 General 2B good for low hardware 6
Gemma3 General, RAG low hardware/1 GPU, context length, multilingual, multimodal
GPT-oss-20b General, Reasoning, Agentic Configurable reasoning
Granite General FOSS, SLM, 4 series low to medium hardware, multi modal
Kimi-K2-Instruct General, Agentic Chat, agentic
Llama3 General 7
Meditron Medical Medical questions, diagnosis, information
Medllama2 Medical Medical questions, trained on open source medical data 8
MedGemma Series Medical Medical text and image comprehension 9
MedImageInsight Medical medical image embeddings (radiology, pathology, etc.) 10
MedImageParse Medical image segmentation 11
CXRReportGen Medical chest X-ray report generation 12
Mistral General, Programming 7B ok for low hardware 13
Moondream Vision Small for edge devices
Nemotron-mini Role-play, RAG, Function 4b for low hardware 14
Phi3 General, RAG low hardware, context length
phi-4 General low hardware, reasoning
Phi4-mini General, RAG low hardware, multilingual, context length
Qwen3 series General Multiple models depending on use
Qwen-Image Image generation, editing Good at text, especially Chinese
StarCoder Programming Trained on 80+ languages, Small to large models 15
WizardCoder Programming 16
Zephyr Assistant Trained version of Mistral and Mixtral as help assistant 17

multimodal means the model can do text and image

Specific Models

Section titled “Specific Models”
Model Name Use Cases Strengths
SimulStreaming (replace Whisper) Speech to text, translation
Voxtral-Mini-4B-Realtime-2602 Speech to text low hardware

Footnotes and Sources

Section titled “Footnotes and Sources”

Model and RAM Recommendations

Section titled “Model and RAM Recommendations”
Model Size RAM
7B 8 GB
13B 16 GB
33B 32 GB

From Ollama README Guidance 18

Models for Programming, Computer Language Development and Use Cases

Section titled “Models for Programming, Computer Language Development and Use Cases”

Summary

Section titled “Summary”

Source: Choosing the right model in GitHub Copilot - Microsoft Tech Community

General Development Tasks

Section titled “General Development Tasks”

New functions, creating tests/documentation, improving code

  • GPT-4.1
  • GPT-5-mini
  • Claude Sonnet
  • Big Pickle (OpenCode Zen)

Light Tasks

Section titled “Light Tasks”

Quick explanations, JSON/YAML transformations, small refactors, regex creation, short Q&A

  • Claude Haiku 4.5
  • MiMo V2 Omni (multi-modal)
  • Nemotron 3 Super Free (text only)

Complex Debugging, Deep Reasoning

Section titled “Complex Debugging, Deep Reasoning”

Analyzing code, debugging hard issues, architecture decisions, multi-step reasoning, performance analysis

  • GPT-5-mini
  • GPT-5.1
  • GPT-5.2
  • Claude Opus

Multi-step Agentic Development

Section titled “Multi-step Agentic Development”

Entire repository changes, migrations, creating features, multiple file planning, automated workflows (plan > run > change)

  • GPT-5.1-Codex-Max
  • GPT-5.2-Codex
  • MiMo V2 Pro

Specific Use Cases

Section titled “Specific Use Cases”

Source: FY26—Advanced-GitHub-Copilot-Workshop/08-models-context/docs - GitHub with some personal notes

Task Recommended Model Why
General completion, quick question GPT-4o / GPT-4.1 Free, general purpose
Refactoring a single file Claude Haiku 4.5 (0.33×) Fast, inexpensive for focused edits
Multi-file feature implementation Claude Sonnet 4.6 (1×) Follows instructions
Analyse 100k+ token codebase Gemini 3.1 Pro (1×) Large 1M context
Architecture design Claude Opus 4.6 (3×) Reasoning
Test generation Claude Sonnet 4.6 (1×) Reliable naming + Arrange/Act/Assert structure
SQL query generation GPT-4o (0×) or Sonnet Both handle SQL well; 0× for routine queries
Legacy code explanation Claude Sonnet 4.6 (1×) Superior contextual narrative explanation
Change, pull request summary GPT-5 mini (0×) Fast, accurate, free
Documentation GPT-5 mini (0×) Fast, accurate, free

Older Models

Section titled “Older Models”
Model Name Use Cases Strengths Notes
Qwen2.5-Instruct General long context tasks due to 1 million token context window
Qwen2.5 General 3b for low hardware 19
Qwen2 General, Programming Chat, Small to Large models 20

See also

Section titled “See also”

Resources

Section titled “Resources”

Footnotes

Section titled “Footnotes”

  1. Coding LLMs Copilot Alternatives ↩

  2. Coding LLMs Copilot Alternatives ↩

  3. Local LLMs on Linux with Ollama ↩

  4. Local LLMs on Linux with Ollama ↩

  5. Coding LLMs Copilot Alternatives ↩

  6. I Ran 9 Popular LLMs on Raspberry Pi 5; Here’s What I Found ↩

  7. Local LLMs on Linux with Ollama ↩

  8. 5 easy ways to run an LLM locally ↩

  9. MedGemma supports image understanding across radiology, pathology, dermatology, and others. Available on Google Vertex AI and MedGemma - Ollama ↩

  10. Healthcare AI foundation models (classic) - Microsoft Foundry (classic) portal | Microsoft Learn ↩

  11. Healthcare AI foundation models (classic) - Microsoft Foundry (classic) portal | Microsoft Learn ↩

  12. Healthcare AI foundation models (classic) - Microsoft Foundry (classic) portal | Microsoft Learn ↩

  13. Coding LLMs Copilot Alternatives ↩

  14. I Ran 9 Popular LLMs on Raspberry Pi 5; Here’s What I Found ↩

  15. Tabby ML Windows Install ↩

  16. Coding LLMs Copilot Alternatives ↩

  17. 5 easy ways to run an LLM locally ↩

  18. Ollama README Guidance on models and RAM ↩

  19. I Ran 9 Popular LLMs on Raspberry Pi 5; Here’s What I Found ↩

  20. Tabby ML Windows Install ↩