Cost & Model Choice¶
How much does it cost?¶
Vibe Seller itself is free and open source. The only thing you pay for is your own LLM API key, billed by the provider directly — no middleman.
Rough scale: one ad-tuning run on DeepSeek, 10 campaigns or fewer, costs about 1 RMB (~0.15 USD) and takes 3–5 minutes. More campaigns and more pages mean more tokens. To spend less, run the routine work on a cheap tier and save the flagship models for the hard jobs — see "Which LLM should I pick?" below.
Which LLM should I pick?¶
Pick one that satisfies both: a long context window and vision (multimodal) — GLM-5.3-Flash, for instance, is labelled "1M context" and Vision right in the model dropdown under Settings → AI. The agent reads page screenshots to work out what state a page is in, so a text-only model degrades badly on browser automation; don't miss the Vision / Text only badge in that dropdown.
Then pick the tier by how hard the task is:
| Task type | Tier |
|---|---|
| Complex — exploring a new platform (no built-in Skill), ad optimization, inventory strategy, long multi-page flows | The provider's SOTA flagship |
| Simple — email classification, fixed-flow data extraction, filling forms field by field | A flash / budget tier is enough |
On complex tasks the agent reads its Skill, judges page state, calls tools and takes notes all at once; weaker models skip steps, garble tool arguments or forget earlier decisions. One failed complex run costs more tokens and time than getting it right once on a flagship. For simple tasks the choice barely matters, so save the tokens.
Where to get each key and which preset to choose: Configure AI Models.