Portail

Llama models · Meta

Chat with Llama online, next to every other model.

Meta's open Llama models and its newer Muse line, in one chat with every other lab. No setup, no GPU, no download.

Start freeFree plan · Every model · No card

What Llama is known for

Llama is the open model family that half the industry builds on. The models are quick and cheap to run, which is why a Llama model is Portail's default for new chats. Muse is Meta's newer line.

How Llama works on Portail

Pick a Llama model from the model picker in any chat. It reads the whole conversation, including anything other models said earlier, and your Portail memory. When you want a second opinion, switch models and keep going, or turn on Council to ask several models at once.

Every plan, Free included, can use every Llama model. Plans change the size of your monthly credit allowance, not which models you can pick. Bigger models use credits faster; the cost band in the table shows roughly how fast.

Llama models on Portail

Read from Portail's live model catalog. Cost runs from $ (cheapest) to $$$$.

ModelContextImagesReasoningCost
Muse Spark 1.3 Contributor1MYesYes$
Muse Spark 1.31MYesYes$$$
Muse Spark 1.2 Contributor1MYesYes$
Muse Spark 1.21MYesYes$$$
Muse Spark 1.11MYesYes$$$
Muse Glimmer 30B131KYesYes$$
Llama Guard 4 12B164KYesNo$
Llama 4 Scout1.3MYesNo$
Llama 4 Maverick1MYesNo$
Llama 3.3 70B Instruct131KNoNo$
Llama 3.2 3B Instruct131KNoNo$
Llama 3.2 1B Instruct60KNoNo$

Questions

How many Llama models can I use on Portail?

14 right now, from Meta. The list follows the live catalog, so new Llama models appear as they're added.

Can Llama read images?

9 of the 14 Llama models on Portail accept images. Check the Images column in the table.

Can I use Llama for free?

Yes. The Free plan includes every model, Llama too, with a starter credit allowance and no card. Cheaper Llama models stretch it further.

Do I have to run Llama myself?

No. Portail runs it for you. Pick a Llama model in the list and chat.

Why is Llama a good everyday default?

It's fast and cheap, so it stretches your credits. When a question gets hard, switch the same thread to a bigger model.

Keep reading

Try it on your own prompt.

Every model, one thread, one memory. Free to start.