Checked against the OpenRouter public models feed. Prices are copied, not estimated.

The Best Model ForWhich AI model to use, from ten labs.

Meta (Llama)

Short answer

Meta's current hosted lineup on this feed is the Muse family. The open-weight Llama models are still listed, including Llama 4 Scout and Maverick. Muse Wonder is the frontier hosted model. Llama 4 Scout is the cheap, long-context, open-weight option, with a knowledge cutoff the feed dates to 2024.

Lab site · Filter the full catalogue to Meta (Llama)

What the top models are good for

1. Muse Wonder

Flagship · $4.00 in · $20.00 out · 1,048,576 tokens

Guide note · 19 Nov 2026

OpenRouter describes Muse Wonder as Meta's frontier reasoning model for complex personal and professional agent tasks. It accepts text, images, video, audio, and files, and returns text. No public weights id is listed.

  • Mixed media in one request
  • Frontier slot in Meta's hosted line
  • About a million tokens of context

What it is good for

One request that includes a photo, a clip, and a PDF

The input list is the reason to pay for Wonder. A field report with a photo, a short clip, and the written log is one job instead of three tools.

Muse Wonder is listed at $4.00 input and $20.00 output per million tokens, with a 1,048,576-token context window.

A hard work problem you will check

Use the frontier slot the way you would any flagship: one careful pass, then a person. The feed does not publish a cutoff for Wonder.

Muse Wonder is listed at $4.00 input and $20.00 output per million tokens, with a 1,048,576-token context window.

The tasks are editorial illustrations, not measured benchmarks. The prices in them are read from today's feed.

2. Muse Ember

Budget · $0.10 in · $0.50 out · 1,048,576 tokens

Guide note · 19 Nov 2026

OpenRouter describes Muse Ember as Meta's smallest model for the work agents repeat all day: tool calls, coding subtasks, and routing. It keeps the wide input list — text, image, video, audio, and files — at a much lower rate than Wonder.

  • Routing and repeated tool calls
  • Wide modality list
  • Low listed rate

What it is good for

Route each request to the right specialist

Routing is what the listing names. Keep the menu of routes short and log the ones Ember is unsure about.

Muse Ember is listed at $0.10 input and $0.50 output per million tokens, with a 1,048,576-token context window.

Coding-agent subtasks

A parent agent can hand Ember the narrow step — rename this field, write this test stub — and keep Wonder or another flagship for the plan.

Muse Ember is listed at $0.10 input and $0.50 output per million tokens, with a 1,048,576-token context window.

The tasks are editorial illustrations, not measured benchmarks. The prices in them are read from today's feed.

3. Llama 4 Scout

Open weights · $0.10 in · $0.30 out · 1,310,720 tokens

Guide note · 19 Nov 2026

Llama 4 Scout is the open-weight, long-context Meta model on this feed: a public Hugging Face id, image input, a window above a million tokens, and a low rate. The feed lists a knowledge cutoff of 2024-08-31. It is a reader of documents you supply.

  • Open weights
  • Very long context
  • Low rate and image input

What it is good for

Summarise a 300-page contract on hardware you control

The contract goes in the prompt. Open weights matter when the document cannot leave your machines. The old cutoff does not matter if you are not asking for news.

Llama 4 Scout is listed at $0.10 input and $0.30 output per million tokens, with a 1,310,720-token context window.

Read a photo of a form

Image input plus open weights is the self-hosted document job. Test it on your real forms, which are always worse than a demo. If Scout is thin on that sample, Llama 4 Maverick is the higher-capacity open Llama 4 on the feed and it costs more. That is a step up, not a saving.

Llama 4 Scout is listed at $0.10 input and $0.30 output per million tokens, with a 1,310,720-token context window.

The tasks are editorial illustrations, not measured benchmarks. The prices in them are read from today's feed.

Every current row

18 Meta (Llama) rows on the feed.
ModelKindContextInputOutputListedGuide
Llama 3.1 70B Instruct Chat 131,072 $0.72 $0.72 23 Jul 2024 New, not yet reviewed
Llama 3.1 8B Instruct Chat 131,072 $0.02 $0.04 23 Jul 2024 New, not yet reviewed
Llama 3.2 3B Instruct Chat 131,072 $0.05 $0.33 25 Sep 2024 New, not yet reviewed
Llama 3.3 70B Instruct Chat 131,072 $0.71 $0.71 6 Dec 2024 New, not yet reviewed
Llama 4 Maverick Chat 524,288 $0.35 $1.00 5 Apr 2025 New, not yet reviewed
Llama 4 Scout Chat 1,310,720 $0.10 $0.30 5 Apr 2025 Reviewed
Llama Guard 4 12B Chat 163,840 $0.18 $0.18 30 Apr 2025 New, not yet reviewed
Muse Ember Chat 1,048,576 $0.10 $0.50 9 Nov 2026 Reviewed
Muse Glimmer 30B Chat 131,072 $0.35 $1.50 9 Aug 2026 New, not yet reviewed
Muse Image Image 65,536 $0 $0 26 Aug 2026 New, not yet reviewed
Muse Spark 1.1 Chat 1,048,576 $1.25 $4.25 16 Jul 2026 New, not yet reviewed
Muse Spark 1.2 Chat 1,048,576 $1.25 $4.25 5 Aug 2026 New, not yet reviewed
Muse Spark 1.2 Contributor Chat 1,048,576 $0.10 $0.20 21 Aug 2026 New, not yet reviewed
Muse Spark 1.3 Chat 1,048,576 $1.25 $4.25 2 Sep 2026 New, not yet reviewed
Muse Spark 1.3 Contributor Chat 1,048,576 $0.10 $0.20 2 Sep 2026 New, not yet reviewed
Muse Spark Open Chat 1,048,576 $1.25 $4.25 15 Oct 2026 New, not yet reviewed
Muse Voice Transcribe 1.0 Transcription Unknown $50.00 $0 11 Sep 2026 New, not yet reviewed
Muse Wonder Chat 1,048,576 $4.00 $20.00 12 Oct 2026 Reviewed

Questions

Which Meta (Llama) model should I start with?

For most people, start with Muse Wonder, then drop to the next model on this page when the task is routine. Read the cheaper-model note under each example.

Are the examples benchmarks?

No. They are concrete tasks used to explain fit. The prices next to them are live figures from the feed. The judgment is a guide note.