osFoundry / Platform
Multi Chat
Run the same prompt across two or more models side-by-side. Compare answers, costs, latency — pick the winner.
What it is
Multi Chat runs a single prompt against two or more models at the same time and lays their responses out side by side in one view. Instead of pasting the same question into separate tabs and holding the differences in your head, you send once and read the answers in parallel columns. It lives in the AI & Orchestration part of osFoundry, so the models you compare can be anything you have configured: a local model on your own machine, a cloud model behind your own API key, or a self-hosted endpoint. Because osFoundry is bring-your-own-key, the cost you see for each model is your provider's real rate with no reseller markup on top.
Each column shows the model's answer alongside the numbers that actually decide which one you keep: what the call cost and how long it took to respond. That turns a vague sense of "this one felt better" into a direct comparison of quality, price, and latency for your specific prompt. A cheap fast model may be all you need for routine work, while a slower or pricier one earns its place only on the prompts that are hard. Multi Chat lets you find that line by testing on your own real inputs rather than on a benchmark.
The point is picking the winner for the task in front of you and then moving on with it. Run the comparison, see which model gives the answer you want at a cost and speed you accept, and standardize on it — no guessing, no switching cost to try the next candidate. Because everything runs inside the local-first desktop workspace, your prompts and the responses can stay on your device rather than passing through a third-party middle layer.
Key capabilities
- Send one prompt to two or more models at once and read every answer in parallel columns
- Compare cost per response using your own API key's real rates, with no reseller markup added
- See per-model latency so you know which model is fast enough for the task
- Mix local, cloud, and self-hosted models in the same comparison from one workspace
- Pick the winning model on your own real inputs, not on generic benchmark scores
Use cases
- Developer: test the same coding prompt across a local model and two cloud models to find the cheapest one that still gets it right
- Prompt engineer: iterate on wording and instantly see which model handles the new phrasing best on quality, cost, and speed
- Team lead evaluating providers: run representative prompts side by side to decide which model to standardize the team on before committing
Frequently asked questions
How many models can I compare at once?
Multi Chat runs the same prompt across two or more models side by side. Which models are available depends on what you have configured in osFoundry — local, cloud, or self-hosted — since you bring your own keys.
Where do the cost and latency numbers come from?
Cost is based on your provider's real per-token rate under your own API key, since osFoundry is bring-your-own-key and adds no reseller markup. Latency is measured from the actual response time of each model on your prompt.
Do my prompts get sent to osFoundry's servers?
osFoundry is a local-first desktop workspace, so your prompts and responses can stay on your device. Calls go directly to the model you chose using your own keys; a local model never leaves your machine at all.