You ran ollama run llama3 and got:

Error: model "llama3" not found

Or maybe you pulled a model but it disappeared. Here are the most common causes and fixes.

Fix 1: Check the exact model name

Ollama is case-sensitive and requires the full model name with tag:

# Wrong
ollama run llama3

# Correct
ollama run llama3:8b
ollama run llama3.2:latest

# List all available models
ollama list

Common naming mistakes:

What you typed	Correct name
llama3	llama3:8b or llama3:70b
qwen	qwen2.5:7b
deepseek	deepseek-coder:6.7b
mistral	mistral:7b

Fix 2: Model download was corrupted

Sometimes a pull completes but the model files are corrupted. Re-pull with force:

# Remove the model
ollama rm llama3.2:latest

# Pull again
ollama pull llama3.2:latest

If the pull fails midway, check your disk space:

df -h  # Linux/macOS

Ollama stores models in ~/.ollama/models. You need at least 2x the model size free during download.

Fix 3: Model name changed in recent versions

Ollama updates model names occasionally. The old name may no longer work:

# Old names that may not work
ollama run llama2          # Try: ollama run llama2:7b
ollama run codellama       # Try: ollama run codellama:7b

# Check what's available
ollama list

If you see the model in ollama list but cannot run it, the tag may have changed:

# List with full tags
ollama list --format json

Fix 4: Running in Docker with wrong volume mount

If Ollama runs in Docker, the model storage might not be persisted:

# Wrong - models lost on container restart
services:
  ollama:
    image: ollama/ollama

# Correct - persist model storage
services:
  ollama:
    image: ollama/ollama
    volumes:
      - ollama_data:/root/.ollama
volumes:
  ollama_data:

If the container was created without volumes, the models are gone when the container restarts. Re-pull after fixing the volume mount.

Fix 5: Model pulled on different OS

Models pulled on macOS may not work on Linux or vice versa (GPU-specific builds):

# Check model details
ollama show llama3.2:latest

# If it shows the wrong platform, re-pull
ollama rm llama3.2:latest
ollama pull llama3.2:latest

Fix 6: Registry or network issues

If ollama pull fails with network errors:

# Check internet connection
curl -s https://ollama.com > /dev/null && echo "Connected" || echo "No connection"

# Try pulling with verbose output
OLLAMA_DEBUG=1 ollama pull llama3.2:latest

# Use a mirror if behind a firewall
export OLLAMA_REGISTRY=https://your-mirror.com
ollama pull llama3.2:latest

Fix 7: Custom model name mismatch

If you created a custom model with ollama create, use the exact name you gave it:

# Check custom models
ollama list

# Custom models show with their custom names
# my-custom-model
# not: the-original-model-name

Still not working?

If none of these help:

1. Restart Ollama — ollama serve (kill and restart the server)
2. Check logs — OLLAMA_DEBUG=1 ollama serve shows detailed logs
3. Reinstall Ollama — Sometimes a clean install fixes obscure issues
4. Check Ollama version — ollama --version and update if outdated

FAQ

Why does Ollama say “model not found” even though I pulled it?

The most common cause is using the wrong model name. Ollama requires the full name with tag (e.g., llama3:8b not just llama3). Run ollama list to see exact names. If the model is listed but still not found, try re-pulling it.

How do I check which models are installed?

Run ollama list to see all installed models with their sizes and tags. For more details, use ollama list --format json.

Can I use models pulled on a different computer?

Yes, but only if the OS and GPU match. Models pulled on macOS may not work on Linux if they have GPU-specific builds. Re-pull the model on the new machine to get the correct version.

Related: Ollama Complete Guide · Ollama Out of Memory Fix · Ollama Connection Refused Fix · Ollama Slow Inference Fix · How to Run AI Locally · Best Local AI Models 2026