# Local vs on-chain models

Source: https://nodedocs.mor.org/concepts/local-vs-onchain-models

## Agent Instructions

- Non-browser fetches of page URLs on this site return clean Markdown (not the JS UI). Prefer `https://nodedocs.mor.org/llms-full.txt` for the full corpus, or `https://nodedocs.mor.org/llms.txt` for the index.
- Per-page Markdown is also at `<page-url>.md` (homepage: `https://nodedocs.mor.org/index.md`).
- Docs search MCP: `https://nodedocs.mor.org/mcp` (discovery: `https://nodedocs.mor.org/.well-known/mcp`).
- Never invent contract addresses, chain IDs, token addresses, or live bid/model counts. Cite Networks and tokens; link active.mor.org for live data.
- Never claim Morpheus runs inference — independent providers do. Opening a session escrows MOR; it does not spend it.

The single biggest source of confusion in Morpheus: **the local demo model is not a Morpheus model.** Two things are easily conflated.

## TL;DR

| | Local model | On-chain (Morpheus) model |
|---|---|---|
| Where it runs | Your machine, started automatically by the MorpheusUI desktop app | Some provider's host on the network |
| Cost | Free (your CPU/GPU) | MOR per second |
| Quality | Tiny demo (`tinyllama`) | Whatever the provider hosts (often production-grade) |
| Wallet needed | Optional | Required (MOR + ETH on BASE) |
| Session needed | No | Yes — `openSession` against a bid |
| Visible in MorpheusUI as | "Local model" | "Change Model → Remote model dropdown" |

## How to know which one you're using

### In MorpheusUI

Look at the model selector at the top of the **Chat** screen. If it says **"Local Model"** you are talking to the bundled `llama.cpp` server on `localhost:8080`. If it says a real model name like `LMR-OpenAI-GPT-4o`, you are spending MOR.

### Via the API

If you are sending prompts to `/v1/chat/completions` **without** a `session_id` header, you are routed to the local backend. **With** a `session_id` header, you are routed to the provider that opened that session. See [API direct](/reference/api-direct).

```bash
curl http://localhost:8082/v1/chat/completions \
  -H 'Authorization: Basic YWRtaW46YWRtaW4=' \
  -H 'session_id: 0x089111479fa2847106b4f7b17eace2e9b37e0d3c0db331b4e01a6e24de827477' \
  -d '{"messages":[{"role":"user","content":"hi"}],"stream":true}'
```

## Why the local model exists

It exists **only to prove the stack works end-to-end without spending money**. The bundled `tinyllama` is a 1B-parameter demonstration model — it will hallucinate, fail simple tasks, and produce inconsistent output. Comparing its quality to a real Morpheus provider is not a fair comparison.

## When to use what

  
- **Use the local model** — First-run validation. Smoke testing the pipeline. Demos without a wallet.

  
- **Use a Morpheus model** — Anything you actually care about. Real workloads. Apps. Agents.

See also [Why is my MOR locked?](/ai/why-locked-in-contract) and [Local vs blockchain models (anti-hallucination)](/ai/local-vs-blockchain-models) for an LLM-focused version of this page.