One model gives you a clear explanation. Another writes the email closer to the way you would say it. A multi-model AI Chrome extension lets you choose without keeping a separate chatbot tab for every task. Lensia's sidebar includes a model picker with models from providers including OpenAI, Anthropic, and Google, the companies behind ChatGPT, Claude, and Gemini. You can try different approaches while keeping your work in Chrome. Install Lensia free to explore the models included in its Free plan.
Understand what one sidebar gives you
Lensia provides access to supported models through its own account and plans. That is different from opening your existing ChatGPT, Claude, or Gemini account inside the sidebar. A subscription to one of those services does not transfer its allowance to Lensia.
The models available to you depend on your Lensia tier. Free includes Essential models, Basic adds Advanced models, and Pro adds Frontier models. Check the current picker for names and availability, because the lineup can change.
You can choose a different model without moving the task to another workspace. You might use one model to tighten a paragraph, then select another when an explanation remains unclear. You still need to give each task enough context and check the result.
Having more models is useful when it helps you finish a task. It can also become a distraction if you keep switching after every answer without knowing what you want to improve. Start with a default, then change it for a reason.
Use a default for ordinary tasks
For everyday work, the Auto option can choose a model within the service's tier and tool policy. It keeps the same usage limits as other requests and applies the Auto discount. Auto does not include the Frontier category.
If you prefer a named model, choose one available on your plan and try it on a few common tasks. A short summary, an email rewrite, and a plain-language explanation are enough to see whether its style suits you.
Keep your prompt specific. Choosing a model cannot replace saying which article to summarize, how formal an email should sound, or which part of an explanation confused you. You may get a better result by fixing the prompt before changing the model.
Our Chrome extension buyer's guide suggests a simple test you can reuse when deciding which tools belong in your browser.
Switch when you can name the problem
Maybe the answer is too abstract. Maybe it misses a condition in your document, or keeps producing language that sounds unlike you. State that problem before asking another model to try the task.
For an explanation, try: "Explain the same idea with one concrete example. Assume I understand the vocabulary but not the relationship between the two concepts."
For writing: "Rewrite this using shorter sentences. Keep my opening and the exact request at the end."
Those instructions make the comparison fairer. If you change the model, the prompt, and the source material at once, you cannot tell which change helped. Keep the task constant when you want to compare models, then revise the answer you prefer.
You do not need a ranking of every model for every subject. Your own repeated tasks are a better starting point than a claim that one model is always best.
Compare answers one after the other
- Open Lensia's sidebar and supply the material for your task.
- Choose a model available on your plan, or start with Auto.
- Ask a focused question and read the first answer.
- Identify what is unclear, missing, or unsuitable.
- Select another model in the picker and restate the same task.
- Check both answers against the source or your original draft.
This is a sequential comparison using the model picker. It does not require a screen that runs several models simultaneously. You can copy the two outputs into your own notes if you want to compare them side by side.
Decide which answer helps before asking for a third. Repeated comparisons consume your usage allowance and can leave you with more text to review than the original task required.
Judge reading, writing, and research differently
Reading and document questions
Prefer the answer that preserves the source's meaning and explains what you asked about. Check details in the original document, including the conditions around a recommendation. An attractive summary that drops an exception can be worse than a plainer accurate one.
Use the same attached PDF for each attempt and name the section you care about. Our PDF guide covers attachment limits and focused document questions.
Writing and rewriting
Judge tone, clarity, and whether the facts survived. Two drafts can both be reasonable. The better one is the version you would send after reviewing it, not necessarily the longest or most polished reply.
See our email writer prompts for constraints that preserve your voice and commitments.
Research and factual questions
Model agreement is not proof. Two models can repeat the same unsupported claim. Check the source rather than counting how many answers sound confident. For a changing fact, such as pricing, ask for current evidence and visit the page yourself.
Basic and above include web search and URL fetching. Those capabilities have their own tier gates; selecting a particular model does not unlock a feature outside your plan. Our research guide explains how to keep sourced findings separate from assumptions.
Know when effort and input size matter
More difficult tasks may benefit from a stronger model or more reasoning effort. Lensia's current effort options are Instant on Free, Medium and High on Basic, and Ultra as an additional Pro option. Availability and provider behavior can differ, so use the options shown for your account.
A longer document, a larger conversation, or a more expensive model can consume more of your allowance. Price is not a fixed number of replies. Check your usage rather than assuming that every request costs the same.
Before increasing effort, see whether a smaller question would help. Comparing one section of two reports is easier to inspect than asking for every possible difference. The assistant may also need a clearer file or a missing passage rather than additional reasoning.
Choose a plan after trying your real work
Free gives you a starter allowance for sidebar chat with Essential models and attachments within its limits. Basic offers about ten times the Free allowance and adds Advanced models, search, URL fetching, and Custom Instructions. Pro has the highest allowance and adds Frontier models and opt-in Agentic Mode.
Compare the current plans when a specific feature or allowance limits your work. You do not have to move to the highest tier simply to begin using a sidebar assistant.
Try the Lensia sidebar, choose a default, and switch models when you want another approach.