Cheap and local models

Zeltro doesn’t sell you AI. It runs whichever agent you point it at, against whichever model you want — including free ones running on your own machine.

The default agents bill at frontier-model prices. For a lot of Zeltro work — scaffolding, wiring config, routine edits — a much cheaper model is fine, and the difference is large:

Model	Input / Output per 1M tokens
Claude Sonnet	~$3 / ~$15
Qwen3 Coder Next (OpenRouter)	~$0.11 / ~$0.80
Anything via local Ollama	free

Zeltro already saves tokens before the model is involved: the environment, networking, database and container plumbing are pre-built, so the agent spends its context on your app instead of deriving Docker setup. Switching models compounds that saving rather than replacing it.

How it works

zeltro ai-set --agent <agent> --model <model> --api-base <url> --api-key <key>

--api-base is the important one. Zeltro hands it to the agent through whichever environment variable that CLI reads:

Agent	Variable	Endpoint it expects
`codex`	`OPENAI_BASE_URL`	OpenAI-compatible
`qwen`	`OPENAI_BASE_URL`	OpenAI-compatible
`aider`	`--openai-api-base`	OpenAI-compatible
`claude`	`ANTHROPIC_BASE_URL`	Anthropic-compatible
`gemini`	—	Google account auth only

“OpenAI-compatible” covers OpenRouter, DeepInfra, Together, Fireworks, Ollama, LM Studio and vLLM — nearly everything.

OpenRouter (cheapest hosted)

zeltro ai-set --agent qwen \
  --model qwen/qwen3-coder-next \
  --api-base https://openrouter.ai/api/v1 \
  --api-key sk-or-v1-...

Pay-as-you-go, no subscription. Swap --agent qwen for --agent codex if you prefer Codex’s interface — both read the same variable.

Ollama (free, local, private)

Install Ollama, pull a coding model, point Zeltro at it:

ollama pull qwen2.5-coder:32b

zeltro ai-set --agent qwen \
  --model qwen2.5-coder:32b \
  --api-base http://localhost:11434/v1 \
  --api-key ollama

The key is a placeholder — Ollama ignores it, but the CLIs expect something.

Nothing leaves your machine, which matters for client work under NDA.

Model size matters more than you would like, and this is measured rather than assumed. Tested here against qwen2.5-coder:1.5b, zeltro create --classify-only returned valid, correctly shaped JSON — the plumbing is fine — but recommended a budgeting app for a guitar pedal tracker, with the reason “Laravel is great for building budgeting apps”. Coherent output, incoherent thinking.

So small models fail in the worst way: they succeed mechanically and are wrong on the substance. 32B quantised on 24GB of VRAM is roughly the floor for agentic work; below that, expect plausible nonsense rather than errors.

Qwen Code also requires Node 22 or newer. It installs and runs on Node 20 — npm warns EBADENGINE and it works anyway — but that is unsupported.

LM Studio or vLLM

Both expose an OpenAI-compatible server, so the shape is identical:

zeltro ai-set --agent codex --model <model> \
  --api-base http://localhost:1234/v1 --api-key local

Claude Code against a local model

Claude Code speaks Anthropic’s API shape, not OpenAI’s, so it cannot talk to Ollama directly. Put a translating proxy such as LiteLLM in front:

zeltro ai-set --agent claude --api-base http://localhost:4000

Worth it only if you specifically want Claude Code’s interface over a local model. Otherwise qwen or codex is less machinery.

Going back

zeltro ai-set --agent claude --api-base none --api-key ""

none clears the endpoint; an empty --api-key clears a stored key. Check where you stand with zeltro ai-set --json-output.

Choosing honestly

Cheaper models are genuinely worse at long autonomous work. The realistic split:

Routine edits, scaffolding, config, boilerplate — a cheap or local model is fine and the savings are real.
zeltro create building a whole app from a sentence, or debugging something subtle — frontier models still earn their price.

Nothing stops you switching per task. zeltro ai-set takes a second and the setting is global rather than per-project.