Our current starting point
ChatGPT for the best overall results. Gemini for lower estimated cost.
You can switch providers before a run, so this choice is easy to revisit as models and pricing change.
Quick recommendation
There are two strong places to start.
Choose ChatGPT for accuracy, speed, and consistent output.
ChatGPT produced our best current balance of reliable identification, conservative handling of uncertainty, clean descriptions, and fast runs.
Create an OpenAI API keyChoose Gemini for strong drafts at a lower estimated cost.
Gemini remained a strong option in our current tests, though uncertain identifications and condition choices sometimes required closer review.
Create a Gemini API keyKimi remains available as another visual-reasoning option. Claude remains available for comparison, but we do not currently recommend it for Draft Builder. Start with ChatGPT or Gemini unless you specifically want to evaluate Claude and are prepared to review every field closely.
Provider observations
What stood out in Lakeview Lister.
Fast, clean, and dependable
Our best current balance of reliable identification, conservative handling of uncertainty, structured output, natural descriptions, and speed.
Best for: Most sellers starting with Lakeview Lister.
Strong quality at a lower estimated cost
Strong identification, category completion, and descriptions at a lower estimated cost, with closer review advisable on uncertain items.
Best for: Sellers who prioritize estimated cost while retaining strong overall draft quality.
Too inconsistent for our current recommendation
Claude produced our slowest recent Draft Builder run and encountered repeated overload failures. It also returned placeholder identities, an incorrect category, nonsensical item specifics, malformed descriptions, and unsupported condition or completeness claims.
Use only if: You intentionally want to compare Anthropic and will verify the identity, category, condition, description, and every item specific before saving or publishing.
Useful identification, uneven listing copy
Kimi could identify difficult items, but its descriptions needed more rewriting and it did not match Gemini or ChatGPT on overall value in our current tests.
Best for: Sellers who want to compare another visual reasoning option.
Claude is still selectable during beta, but “supported” does not mean it performed equally well in our testing. Its recent failures can require substantially more review than ChatGPT or Gemini and may be easy to miss if a draft is accepted without checking every field.
AI models, prices, rate limits, and provider behavior change frequently. These observations describe Lakeview Lister testing through July 2026, not every item or future model release.
Current model lineup
Know which model the app will use.
Draft Builder uses the Standard model. Fast ID lets you choose Standard or Economy before a run.
| Provider | Draft Builder & Standard Fast ID | Economy Fast ID |
|---|---|---|
| ChatGPT | GPT-5.6 Terra | GPT-5.6 Luna |
| Claude | Claude Sonnet 5 | Claude Haiku 4.5 |
| Gemini | Gemini 3.1 Pro Preview | Gemini 3.6 Flash |
| Kimi | Kimi K3 | Kimi K3 with lower reasoning |
You add a provider API key; Lakeview Lister supplies the model name and connection settings. The provider bills your account directly for the tokens used.
Choose by workflow
Match the model to the job.
Use the Standard model for listing-ready drafts.
Draft Builder has more work to do: identify the item, organize listing fields, choose an eBay category, and write the description. The added model capability is worthwhile here.
Start with Standard; use Economy for straightforward items.
Standard is the safer default for uncommon or ambiguous items. Economy can reduce AI cost when you mainly need quick identification and can tolerate a little more review.
Photo count, image size, item difficulty, output length, provider pricing, and retries can all change the cost of a run. Lakeview Lister records usage so you can compare your own results.
AI performs more consistently when a Draft Builder run contains similar items with accurate shared defaults. See our Draft Builder best practices before comparing providers on a large batch.
Human review still matters
Review every draft before it reaches eBay.
AI providers do not have dependable access to complete, current sold-listing data. Lakeview Lister’s estimates are primarily informed by currently listed items, which show asking prices—not what buyers actually paid. Always review comparable sold listings before setting your price.
- Confirm the identification, title, category, condition, and item specifics against the item in front of you.
- Verify measurements and included parts instead of assuming the model read every photo correctly.
- Remove any claim about testing, operation, or completeness that you did not personally establish.
- Use the sold-listings link to research price. AI price estimates do not have dependable access to complete, current sales data.
An AI model cannot know whether an item was tested or works correctly from pictures alone. Add those facts only when you know them yourself.
How we tested
Same app, same item groups, different providers.
We compared providers inside Lakeview Lister using repeated photo groups and the same Draft Builder settings, then reviewed identification, category selection, item specifics, description quality, speed, and estimated cost. We also compared Standard and Economy Fast ID runs.
This was practical product testing, not a scientific benchmark. Results can change with the item, photo quality, provider response, model update, and prompt improvements in a newer Lakeview Lister build.