Vendor · Language models · Image models · Video models · Specialist models
Meta has 32 models in the directory. The strongest is Muse Wonder Full: Rank 6 to 9 of 95 in the overall index, with 2 models among the best ten.
Llama as an open-weight model family, with Meta AI in WhatsApp, Instagram, and wearables. Meta acquired Limitless. Limitless stopped selling its Pendant and said it would support existing customers through 2026.
In the directory: 29 language models, 1 image model, 1 video model and 1 specialist model. 26 of them with open weights that can run locally.
Every model there ever was. Models with a profile lead to every benchmark, price and source.
| Muse Spark Open | Oct 15, 2026 | Active | n/a | n/a | 170 | 1.05M | 43 |
| Muse Wonder | Oct 12, 2026 | Preview | 83Rank 9 to 12 of 95 | $4 / $20 | 197 | 1.05M | 3 |
| Muse Wonder Full | Oct 12, 2026 | Subscription only | 90Rank 6 to 9 of 95 | n/a | n/a | 1M | 16 |
| Muse Spark 1.3 | Sep 2, 2026 | Active | 69Rank 13 to 19 of 95 | $1.25 / $4.25 | 198 | 1.05M | 50 |
| Muse Glimmer 30B | Aug 10, 2026 | Open source | n/a | $0.30 / $1.20 | 125 | 131K | 45 |
| Muse Spark 1.2no profile | Aug 5, 2026 | Retired | n/a | n/a | n/a | n/a | n/a |
| Llama 4 Maverick 17B Instruct | Apr 5, 2025 | Active | n/a | n/a | n/a | 1M | 32 |
| Llama 4 Scout 17B Instruct | Apr 5, 2025 | Active | n/a | n/a | n/a | 10M | 35 |
| Llama 4 Behemothno profile | April 2025 | No status data | n/a | n/a | n/a | n/a | n/a |
| Llama 4 Maverick | April 2025 | Active | below 1 | $0.35 / $1 | 54 | 1.05M | 45 |
| Llama 4 Scout | April 2025 | Active | below 1Rank 95 of 95 | $0.10 / $0.30 | 67 | 10.49M | 24 |
| Llama-3.3-70B-Instruct | Dec 6, 2024 | Active | n/a | n/a | 88 | 128K | 50 |
| Llama 3.3 70B | December 2024 | Active | n/a | $0.45 / $0.90 | 89 | 131K | 50 |
| Llama 3.2 1B | Sep 25, 2024 | Active | n/a | $0.027 / $0.20 | n/a | 131K | 41 |
| Llama 3.2 1B Instruct | Sep 25, 2024 | Active | n/a | n/a | n/a | 131K | below 1 |
| Llama 3.2 3B | Sep 25, 2024 | Active | n/a | $0.05 / $0.33 | n/a | 131K | 42 |
| Llama 3.2 3B Instruct | Sep 25, 2024 | Active | n/a | n/a | n/a | 131K | 8 |
| Llama-3.2-11B-Vision-Instruct | Sep 25, 2024 | Active | n/a | n/a | 29 | 128K | 41 |
| Llama-3.2-90B-Vision-Instruct | Sep 25, 2024 | Active | n/a | n/a | n/a | 128K | 29 |
| Llama 3.1 405B Instruct | July 2024 | Active | n/a | n/a | n/a | 131K | 33 |
| Llama 3.1 70B | July 2024 | Active | n/a | $0.40 / $0.40 | 63 | 131K | 19 |
| Llama 3.1 70B Instruct | July 2024 | Active | n/a | n/a | n/a | 131K | 35 |
| Llama 3.1 8B | July 2024 | Active | n/a | $0.05 / $0.08 | 141 | 131K | 28 |
| Llama 3.1 8B Instruct | July 2024 | Active | n/a | n/a | n/a | 131K | 47 |
| Llama 3 70Bno profile | April 2024 | No status data | n/a | n/a | n/a | n/a | n/a |
| Llama 3 8Bno profile | April 2024 | No status data | n/a | n/a | n/a | n/a | n/a |
| Code Llama 70B Instruct | January 2024 | Active | n/a | n/a | n/a | n/a | 23 |
| Llama 2 70Bno profile | July 2023 | No status data | n/a | n/a | n/a | n/a | n/a |
| OPT-175Bno profile | May 2022 | No status data | n/a | n/a | n/a | n/a | n/a |
| Muse Image | Jul 7, 2026 | Active | $0.01 per image | 33 |
| Muse Video | n/a | Preview | n/a | 62 |
| SAM 3 | n/a | Limited access | n/a | 68 |
Language model · Oct 15, 2026
Muse Spark Open is Meta's open variant of Muse Spark, with 816 billion total and 42 billion active parameters. The model is natively multimodal, offers 1,048,576 tokens of context and targets coding and agentic work.
Language model · Sep 2, 2026
Muse Spark 1.3 is Meta’s proprietary reasoning model for long agentic and coding tasks. Meta publishes no checkpoint for local deployment.
Language model · Aug 10, 2026
Muse Glimmer 30B is Meta’s open agent model for local use. Unlike Muse Spark, the focus is a downloadable checkpoint that combines image understanding and controllable reasoning in a compact architecture.
Image model · Jul 7, 2026
Muse Image’s defining feature is Meta’s concrete image agent. It can call search and coding tools, refine its own drafts, and spend more computation at test time. Meta links the coding step to more accurate charts and QR codes, while multiple reference images can remain consistent across editing turns.
Language model · Apr 5, 2025
Llama 4 Maverick 17B Instruct is the cataloged Instruct identity for Meta’s open Maverick weights. The 17B in the name describes active parameters, not the full size of the MoE model. It accepts text and images and provides about one million tokens of context.
Language model · Apr 5, 2025
Llama 4 Scout 17B Instruct is the cataloged Instruct identity for Meta’s open Scout weights. As with Maverick, 17B describes active rather than all MoE parameters. Its defining property is a context window of about ten million tokens for text and images.
Language model · April 2025
Llama 4 Maverick brought a natively multimodal mixture-of-experts model to Meta’s open Llama line. Its weights are available under the Llama 4 Community License, so the checkpoint can be run independently.
Language model · April 2025
Llama 4 Scout is the MoE variant in the first Llama 4 generation with a context window of roughly ten million tokens. It processes text and images.
Language model · Dec 6, 2024
Llama 3.3 70B Instruct is the open Instruct checkpoint of the Llama 3.3 70B generation. The model card lists tool calling and the Llama 3.3 Community License. Llama 3.3 70B Instruct is documented as a self-hostable weight checkpoint.
Language model · December 2024
Llama 3.3 70B is Meta’s later dense 70B checkpoint for text and code. Compared with Llama 3.1 70B, it keeps the same broad approach but belongs to the subsequent 3.3 generation.
Language model · Sep 25, 2024
Llama 3.2 1B was built through pruning and knowledge distillation. Meta used token probabilities from Llama 3.1 8B and 70B as training targets to recover performance after reducing the model. It is the smallest text checkpoint in the Llama 3.2 line.
Language model · Sep 25, 2024
Llama 3.2 1B Instruct is the instruction-tuned version of the smallest Llama 3.2 model, with roughly 1.2 billion parameters. Compared with the base model Llama 3.2 1B, it is aligned for dialogue through supervised fine-tuning and RLHF.
10 more current models are in the model list above.
Derived from the overall index, the API price and the measured speed of the current models.
The strongest choice
Muse Wonder Full
The vendor's highest overall index
The fastest
Muse Spark 1.3
198 tokens/s, a typical model manages 98
The 20 best language models, with the models from Meta highlighted. “All” and the filter show every model in the index.
20 of 95 models
Provider
Weights
Price (blended per 1M tokens)
Released
Anthropic
OpenAI
Meta
Moonshot AI
DeepSeek
xAI
Z.ai
Xiaomi
Alibaba
StepFun
Mistral AI
MiniMax
Thinking Machines Lab
Inception
| Vendor | Models | Model types | Best language model | Top 10 | Open weights | New in 2026 |
|---|---|---|---|---|---|---|
| Anthropic | 33 | 33 language models | Claude Opus 5.5 (Rank 1 to 3) | 4 | 0 | 13 |
| OpenAI | 63 | 46 language models, 4 image models, 2 video models, 9 audio models, 2 specialist models | GPT-6 Astra (Rank 4 to 8) | 2 | 4 | 27 |
| 66 | 43 language models, 4 image models, 6 video models, 11 audio models, 2 specialist models | Gemini 4 Carbon (Rank 1 to 5) | 2 | 19 | 26 | |
| Meta | 32 | 29 language models, 1 image model, 1 video model, 1 specialist model | Muse Wonder Full (Rank 6 to 9) | 2 | 26 | 7 |
| DeepSeek | 23 | 23 language models | DeepSeek V4.1 Flash 1017 (Rank 13 to 20) | 0 | 23 | 9 |
| Moonshot AI | 12 | 12 language models | Kimi K3.1 Pro (Rank 10 to 13) | 0 | 10 | 7 |
| xAI | 16 | 12 language models, 2 image models, 2 video models | Grok 4.7 (Rank 18 to 24) | 0 | 1 | 12 |
By release date. Grey: retired models.
7 releases since January 2026, by release date.
Prices per 1M tokens for the current language models, according to Meta or the provider linked in each model profile.
| Current model | Input | Output | Context |
|---|---|---|---|
| Muse WonderPreview | $4 | $20 | 1.05M |
| Muse Spark 1.3 | $1.25 | $4.25 | 1.05M |
| Llama 4 Maverick | $0.35 | $1 | 1.05M |
| Llama 4 Scout | $0.10 | $0.30 | 10.49M |
| Llama 3.2 1B | $0.027 | $0.20 | 131K |
| Llama 3.2 3B | $0.05 | $0.33 | 131K |
| Llama 3.1 8B | $0.05 | $0.08 | 131K |
| Muse Glimmer 30B | $0.30 | $1.20 | 131K |
| Llama 3.1 70B | $0.40 | $0.40 | 131K |
| Llama 3.3 70B | $0.45 | $0.90 | 131K |
| Current model | Price |
|---|---|
| Muse Image | $0.01 per image |
Monthly prices in Germany incl. VAT; “≈” is converted from US dollars. The vendor's models are often also part of other providers' plans, see the AI subscription comparison.
| Plan | Per month | Note |
|---|---|---|
| Meta AI Free | free |
The directory lists 32 Llama models from Meta: 29 language models, 1 image model, 1 video model and 1 specialist model. It counts every model Meta ever released, including retired ones; 26 of them have their own profile. 22 are currently available, the newest being Muse Spark Open, Muse Spark 1.3 and Muse Glimmer 30B.
Muse Wonder Full. It is at rank 6 to 9 of 95 in the overall index, which combines independent benchmarks from several sources.
Muse Spark Open, released on Oct 15, 2026. The timeline below shows every release by date.
Through the API per 1M tokens (input / output): Llama 3.2 1B $0.027 / $0.20. The prices of all 10 current language models are in the price table.
Yes, with the Meta AI Free plan.
In the overall index, Muse Wonder Full (rank 6 to 9 of 95) and the best OpenAI model, GPT-6 Astra (rank 4 to 8 of 95), are about level. The gap is too small for a reliable call.
Llama is developed by Meta. The company is based in Menlo Park and was founded in 2004.
Yes, 26 of the 32 models have open weights. Each model profile lists how much memory it needs.
Muse Spark 1.3 with 198 tokens per second according to Artificial Analysis. A typical model manages 98.
The largest is Llama 4 Scout with 10.49 M tokens (≈ 15,730 A4 pages). Every model's value is in the model list.
Data as of Nov 1, 2026