Text model

LLaMA and Muse Spark — Meta models

Q: Which Meta models are available in GPTunneL?

The current list is always visible in the «Every Meta model» block above — it is built straight from the platform catalogue.

Q: How is Muse Spark different from LLaMA?

Muse Spark is Meta's closed flagship: the weights are not published, but it is noticeably stronger on code and agentic runs and holds a million tokens of context. LLaMA stays the open line — you can deploy it yourself, and it costs less.

Q: What does «open model» mean?

The weights are published: you can download the model and run it on your own server rather than only calling someone else's API. In GPTunneL the same models are available without any hardware of your own.

Q: How is LLaMA different from ChatGPT?

LLaMA is open and noticeably cheaper, but it trails the newest closed flagships on hard reasoning. People take it when price matters or when they need a model they can deploy themselves.

Q: Which languages does LLaMA handle?

The family is multilingual and answers confidently in dozens of languages, though it trails ChatGPT and Claude on the finer points of style.

Meta models in GPTunneL: the Muse Spark 1.3 flagship with a 1M token context and the open LLaMA line. Pay per token, no subscription and no VPN.

Try Muse Spark Available models

What Meta's models are good at

Traits Meta keeps from one generation to the next — you can count on them whichever model you pick.

Open weights

The defining difference of the LLaMA line: the models are published and can run on your own server. Neither ChatGPT nor Claude offers that.

Low price

Openness creates competition among the providers that run these models — and keeps the price noticeably below closed equivalents.

The Muse Spark 1.3 flagship

Meta's top model with a million-token context: hard code, long agentic runs and whole folders of documents in one pass.

The base for fine-tuning

Most open models on the market grew out of LLaMA: it has more ready fine-tunes and tooling around it than anything else.

Which model to pick

Meta's flagship and two open fourth-generation models — for different jobs and budgets.

	Muse Spark 1.3Code and agents	ScoutThe default pick	MaverickHard tasks
Best for	Hard code, agentic runs, long context	Everyday writing, correspondence, simple code	Hard reasoning, analysis, large documents
Context	1M tokens	1M tokens	1M tokens
Speed	Medium	High	Medium
Price	Above the open models, below the flagships	Low	Medium

Figures come from the platform catalogue. Exact per-model prices are on the Pricing

Every Meta model

Release dates follow Meta's official announcements.

September 2026

Muse Spark 1.3

Meta's closed flagship, sitting next to the open line: a million-token context, text, image and video input, and a focus on code and long agentic runs. The xhigh version everyone can use scores 61 on the Artificial Analysis Intelligence Index — level with GPT-5.6 Sol and Grok 4.6.

April 2025

LLaMA 4 Scout LLaMA 4 Maverick

The first generation with a million tokens of context. Scout is the everyday workhorse, Maverick handles the heavier jobs.

December 2024

LLaMA 3.3 70B

A 70-billion-parameter model that matched the quality of earlier flagships three times its size.

July — September 2024

LLaMA 3.1LLaMA 3.2

Context grew to 128 thousand tokens, very compact versions appeared, and so did the first models that understood images.

April 2024

LLaMA 3

The generation where open models were first seriously compared with closed flagships.

Full release history

July 2023

LLaMA 2

The first version licensed for commercial use. The whole open-model ecosystem starts here.

February 2023

LLaMA 1

The research release that began the history of open large language models.

How billing works

There is no subscription: you top up one balance and spend it on any model on the platform.

Pay per token

You are charged for exactly what the request and the answer used. Not using it costs nothing: no limits, no monthly fee.

One balance for every model

LLaMA, ChatGPT, Claude, image and video generation — all out of the same wallet. No separate subscription per service.

Top up the way you prefer

International cards, Apple Pay, Google Pay or crypto. The minimum top-up is $5.

See pricing

How to get started with Meta's models

One account for every model on the platform.

Emailyou@example.comPassword••••••••

Draft an outline for an article about a company moving to remote work

LLaMA 4 Scout

Article outline

Open with the numbers: how many people moved, over what period, and what happened to productivity.

The main section covers the three difficulties they hit and what worked against each one.

Close with a checklist for companies that are only now considering the move.

The model switches right inside the input — you can change it mid-conversation.

The answer lands in the chat and the context is kept until the conversation ends.

Try Meta's models in GPTunneL

Signing up takes a minute, and $5 on the balance is enough to see whether the model fits your task.

Try Muse Spark

Frequently asked questions

/01Do I need a VPN to use Meta's models through GPTunneL?

No. Requests go through GPTunneL's infrastructure, so the models open like any ordinary website — no VPN and no proxy.

/02Which Meta models are available in GPTunneL?

/03How is Muse Spark different from LLaMA?

/04What does «open model» mean?

/05How is LLaMA different from ChatGPT?

/06Which languages does LLaMA handle?

LLaMA is Meta's family of open neural networks and the foundation of almost the entire open-model ecosystem: most free language models on the market grew out of it in one way or another.

The key difference from ChatGPT and Claude is the published weights. The model can be downloaded and run on your own server, which means nobody can switch it off or change it without warning. Price follows from that openness: anyone can run LLaMA, and the competition keeps the cost well below closed equivalents. The larger fourth-generation models hold a million tokens in context.

In the autumn of 2026 the family gained an older sibling — Muse Spark 1.3. Meta does not publish its weights: this is a closed flagship for hard code and long agentic runs, with a million-token context and image and video input. The full breakdown is in the Muse Spark 1.3 article.

Meta models are available in GPTunneL without a server of your own — worldwide, without a VPN, with one balance and billing for the tokens you actually spend.