Pricing¶
/pricing is the LLM pricing browser. It surfaces the cached
per-(provider, model) rates AMX uses to compute USD cost on every run.
Where the numbers come from¶
AMX fetches rates from two sources:
- LiteLLM — community-maintained pricing catalogue, broad coverage
- OpenRouter — authoritative for OpenRouter-routed models
Rates are cached on disk with a freshness timestamp. The top-bar pricing-cache badge turns warm-yellow when the cache is more than 24 hours old; clicking it forces a refresh.
A bundled fallback ships with AMX so a brand-new install with no network access still has reasonable estimates.
Table¶
| Column | Notes |
|---|---|
| Model | model_id, mono font |
| Provider | Provider hint |
| Input $/Mtok | Per-million-token input price |
| Output $/Mtok | Per-million-token output price |
| Source | litellm / openrouter / fallback |
| Fetched | Relative age of the cached price ("2 h ago") |
The table is searchable (by model ID, provider, source) and filterable by source. Sort by clicking any column header.
Staleness banner¶
When the cache is older than 24 hours, a yellow banner appears above the table:
These prices are more than 24 h old. Click the refresh icon in the topbar to pull fresh data.
The same staleness signal drives the topbar badge colour and the cost columns on Overview (no warning, but the totals are pinned to the rates that were live when each run ran).
Empty state¶
A brand-new install with no cache and no network falls back to the bundled rates. The empty state renders:
No price data loaded yet. Click the refresh icon in the topbar or check that the AMX process can reach the public LiteLLM / OpenRouter endpoints.
CLI equivalents¶
| Studio | CLI |
|---|---|
| Pricing table | /refresh-prices (re-fetches into the same cache) |
| Per-model override | /cost <input> <output> on the active profile |