Local model choice can turn into name collecting very quickly. Llama, Qwen, Gemma, Mistral, DeepSeek, Phi, gpt-oss, coder models, reasoning models, small models, giant models, quantized models, preview models: it is a lot.

The goal is not to memorize every release. The goal is to choose a model with a reason.

Mental model: model choice is task fit plus hardware fit

A useful model is the intersection of three things:

QuestionWhy it matters
What task do I need?Chat, coding, summarization, research, structured extraction, multilingual work, or tool use require different strengths
What hardware do I have?A model that barely fits may be less useful than a smaller model that runs smoothly
What constraints matter?Privacy, license, cost, speed, context length, and reliability change the decision

Model Selection Matrix

Click a pillar to see what happens when it's ignored

Task Fit
Coding, Chat, RAG
Hardware Fit
VRAM, RAM, CPU
Constraints
Privacy, Speed, Cost
The "best" model on a leaderboard may be the wrong model for your laptop.

Common model families

As of May 2026, these are common families beginners will encounter:

FamilyWhy people use itWhat to check
LlamaStrong general open model ecosystem, many variants and quantizationsLicense, exact generation, context length, tool support
QwenStrong multilingual and coding options, many sizesPrompt format, reasoning behavior, quantization availability
GemmaCompact Google model family with accessible sizesLicense, local runtime support, task fit
MistralEfficient models and strong European open model ecosystemInstruct format, license, model age
DeepSeekPopular for reasoning and coding variantsDistilled vs original variants, runtime behavior
PhiSmall models that can be practical on modest hardwareWhether the small size is enough for your task
gpt-ossOpen-weight OpenAI model family for local or self-hosted useOfficial release notes, license, recommended runtime
Coder modelsSpecialized models for code completion, refactoring, and repository tasksLanguage support, tool calling, repo-context handling

This table is a map, not a ranking. Model families evolve, and each family contains multiple checkpoints.

Size is not intelligence

Model size matters, but it is not the only thing that matters. A larger model often knows more patterns and handles harder tasks better, but it also needs more memory and compute. A smaller model can be better if it responds quickly, fits the task, and is paired with good context.

Watch for labels such as:

LabelMeaning
3B, 7B, 8B, 14B, 32B, 70BApproximate parameter scale
instructTuned to follow instructions
baseRawer pretrained model, usually not ideal for chat
coderTuned or optimized for programming tasks
reasoningTuned to spend more effort on complex problems
visionCan accept image inputs
tool useCan produce structured calls to tools

For beginners, an instruct or chat model is usually the right starting point.

Evaluation beats vibes

Instead of asking which model is best, create a tiny local evaluation set:

  1. A short writing task you actually care about.
  2. A factual explanation task where you know the answer.
  3. A coding or technical task from your real work.
  4. A summarization task using your own text.
  5. A structured output task, such as producing JSON or a table.

Run the same tasks across two or three candidate models. Score them with simple labels: good, usable, weak, failed. Add notes on speed and memory.

Key papers on model evaluation and selection

What about llmfit?

Tools such as llmfit are meant to help answer “what can run on this machine?” by inspecting hardware and ranking candidate local models. That is a useful category of tool, but do not treat any helper as an oracle.

Before recommending or installing a newer CLI, verify:

CheckWhy
Official repositoryAvoid package-name confusion
Package publisherConfirm it matches the project
Release historyNew projects can change quickly
Install methodPrefer official docs over copied commands
Local data behaviorHardware scanners may collect system details

Use fit tools to narrow the search, then test the final candidates yourself.

A practical model selection workflow

  1. Define the task: chat, coding, research, summarization, extraction, or agent use.
  2. Write down hardware: RAM, VRAM, CPU/GPU, operating system.
  3. Pick a size class that should comfortably fit.
  4. Choose two model families, not ten.
  5. Start with instruct variants.
  6. Prefer popular quantizations with active runtime support.
  7. Run your tiny evaluation set.
  8. Keep the faster model if quality is close.

What to verify because models change

Verify the current model list, license, context length, supported parameters, and runtime compatibility at drafting time. For hosted endpoints, check the provider’s metadata, not only the model card. For local files, check the exact checkpoint and quantization.

Useful places to verify:

The most reliable model choice is not glamorous. It is the smallest model that does your real task well enough, fast enough, and within your constraints.