Aider
Git-aware AI pair programmer in the terminal.
- Website
- https://aider.chat
- Pricing
- Free and open source (Apache 2.0); bring your own model API key or local model. Price checked 2026-10-14 source
- Affiliate program
- none (OSS). Link above is a plain vendor URL.
Rankings
| Use case | Rank | Score | Status |
|---|---|---|---|
| Best terminal / CLI coding agent | 6/6 | 70 | Partly hands-on tested 2026-10-14 |
| Best free / open-source / bring-your-own-key | 2/6 | 77 | Preliminary · based on public info |
Hands-on test log
2026-10-14 · Best terminal / CLI coding agent · cli-agent fixed tasks v1 · Partly hands-on tested 2026-10-14
Environment: aider 0.86.2 on Linux, model qwen2.5-coder-3b-instruct Q4_K_M running locally via Ollama 0.40.4 on CPU ($0 cost). One non-interactive run per task, no human edits.
Command: aider --model ollama_chat/qwen-coder-3b --no-auto-commits --yes-always --no-stream <files> --message "<task prompt>"
| Task | Result | Notes |
|---|---|---|
| T1 fix failing test | PASS | Changed range(len(xs) - n) to range(len(xs) - n + 1) and added an n > len(xs) guard; 2/2 tests pass; 26 s |
| T2 implement slugify from tests | FAIL | Implemented slugify correctly but also added two incorrect assertions to the test file, despite being told not to modify tests; pytest fails; 22 s |
| T3 multi-file rename | PASS | Renamed calc_total to compute_total in cart.py and checkout.py; 3/3 tests pass; 32 s |
2 of 3 tasks passed with a small local 3B model. This checks Aider's workflow (repo map, edit format, file edits). It does not show Aider's best case with a frontier model. The score stays preliminary until the full suite runs with the reference model.
Raw logs: t1.log, t2.log, t3.log · task definitions
Alternatives to Aider
Updates
Spotted an error? Send a correction.