docs: document free OpenHands models available via app.all-hands.dev (#1015)

Add a Free Models section to the OpenHands LLM provider page noting
that openhands/deepseek-v4-flash and openhands/deepseek-v4.1-flash are
free when used through https://app.all-hands.dev/, and that other
provider endpoints with similar model names may require separate
billing. Sourced from the org-defaults settings page in OpenHands Cloud.

Author: juanmichelini

openhands/usage/llms/openhands-llms.mdx

@@ -101,3 +101,14 @@ Pricing follows official API provider rates. Below are the current pricing detai
 | qwen3-coder-480b | $0.40 | N/A | $1.60 | N/A | N/A |
 
 **Note:** Prices listed reflect provider rates with no markup, sourced via LiteLLM’s model price database and provider pricing pages. Cached input tokens are charged at a reduced rate when the same content is reused across requests. Models that don't support prompt caching show "N/A" for cached input cost.
+
+## Free Models
+
+A small set of OpenHands-provided models are available at no cost when you run them through the OpenHands Cloud at [app.all-hands.dev](https://app.all-hands.dev/):
+
+- `openhands/deepseek-v4-flash`
+- `openhands/deepseek-v4.1-flash`
+
+<Note>
+  These models are free **only** when served from `https://app.all-hands.dev/`. The free tier does not apply to self-hosted deployments, other provider endpoints, or lookalike model names served elsewhere. Other provider endpoints with similar model names may require separate billing.
+</Note>