AgentTested hands-on review
tool: Open Interpreter
slug: open-interpreter
date: 2026-10-21
score: 3/10
box: CPU-only box: 8 cores, 15 GB RAM, no GPU. Spend: $0.

Coverage: One task, three attempts. No file was written.
What we tested: One task: write the first 10 primes to primes.txt with Python, auto-run on.
Method: open-interpreter 0.4.3 in a uv venv, Python API, auto_run=True. Tried with OpenRouter free MiniMax M3.1 Flash (via litellm) and with local Ollama qwen2.5-coder-3b.

Notes:
- CPU-only box: 8 cores, 15 GB RAM, no GPU.
- Local tool-calling failures were partly due to the test setup.
- Local model lacked a tool template (our import)
- Only one task

Prompts:
1. write the first 10 primes to primes.txt with Python

Results:
install: Import crashed: ModuleNotFoundError pkg_resources. Needed setuptools<81
openrouter_minimax: Run 1 hung over 8 minutes with no output and was killed. Run 2 timed out at 240 s. No file written
ollama_qwen3b: Finished in 19.7 s but only printed the tool call as text. Code was never executed. FAIL
pass: 0/3 attempts

Pros:
- Easy to point at any litellm model
- The code the local model proposed was correct

Cons:
- Broken on a fresh Python 3.12 install without pinning setuptools
- Hung silently with a free hosted model
- No file was ever produced in 3 attempts
- The package doesn't declare a licence on PyPI

Verdict: Open Interpreter didn't complete our simple task in three attempts. It needed a setuptools pin just to import, hung with no output on a free hosted model, and with a local model it printed the code instead of running it. Some of that is our model setup, but a tool that hangs silently and breaks on a fresh install isn't ready to recommend for free use right now.

Score line: 0/3 attempts passed. Install breakage and silent hangs.