ASSISTANT · LLAMA · AGENTS
Meta's assistant across Facebook, Instagram, WhatsApp, plus Llama foundation models + agent partnerships.
Current models
General-purpose agent — Llama 4 + Claude dual-model architecture.
About Meta AI
Meta AI’s Llama family, developed by Meta, comprises multimodal language models that serve as both an assistant and the open‑weight foundation for many local and open‑source applications, positioning them as widely used components within the broader ecosystem of accessible AI tools.
Meta AI sits across Meta's social products + Llama open-weights ecosystem. OSS also tracks adjacent products built on Llama — most notably Manus, the general-purpose agent from Butterfly Effect that runs on Llama 4 + Claude in dual-model mode.
Manus Joins Meta
Manus has joined Meta, bringing its AI capabilities to businesses worldwide. The company will continue to operate from Singapore and sell its product subscription service through its app and website. Manus's solution is currently driving value for millions of users worldwide.
Manus — Blog
Meta Tests User Willingness to Pay for AI
Meta is testing user willingness to pay for its AI technology through Meta One Subscriptions. The test aims to determine how much users are willing to pay for access to Meta's AI features.
Meta AI & Llama — Coverage
Posts to your status feed
Pick the closest match below, edit the body, and post. Your report carries the #meta tag automatically so it surfaces here + in the trending-tags rail. Reporting also follows Meta AI so you’ll get status updates.
Best-effort snapshot of public consumer plans (US$/mo) as of 2026-06. Prices and tiers drift — confirm on the provider's own pricing page before relying on these.
Track your Meta AI plan in My AI Plans →
Free
Free
No paid consumer tier — Meta AI is free in Meta's apps. Llama open weights are free to download (license terms apply); the Llama API is metered.
Manus Joins Meta
Manus has joined Meta, bringing its AI capabilities to businesses worldwide. The company will continue to operate from Singapore and sell its product subscription service through its app and website. Manus's solution is currently driving value for millions of users worldwide.
Manus — Blog
Meta Tests User Willingness to Pay for AI
Meta is testing user willingness to pay for its AI technology through Meta One Subscriptions. The test aims to determine how much users are willing to pay for access to Meta's AI features.
Meta AI & Llama — Coverage
No articles yet in Meta AI. Check back soon, or browse all sections.
No discussions yet.
Be the first to start one — your post will surface here + in every follower’s feed.
Loading edit history…
Llama 4 Maverick is the flagship model in Meta’s Llama series, launched on 1 April 2025 with a knowledge cut‑off matching its release date. It features an expanded context window of one million tokens, allowing it to process substantially longer inputs than earlier Llama versions.
Llama 4 Scout is an open‑weight model from Meta’s Llama series that entered active service on 1 April 2025, with its knowledge base frozen at that date. Compared with earlier Llama releases, Scout provides a substantially larger context window of roughly ten million tokens, enabling it to process far longer inputs while retaining the same open‑weight licensing approach.
Llama 3.3 70B is an active, balanced‑tier model from Meta released in December 2024 with a knowledge cutoff of October 2024 and a 128 000‑token context window. It is positioned as a cost‑effective alternative that delivers text‑only performance comparable to the higher‑parameter Llama 3.1 405B, and it is recommended by the community as the default choice for such tasks.
Llama 4 Maverick released
Flagship MoE model, 17B active / 400B total, natively multimodal. Launch drew LMArena benchmaxxing controversy. Month known, exact day not stated.
Llama 4 Scout released
First Llama MoE models; Scout offers a 10M-token context window. Month known, exact day not stated.
Llama 3.3 70B released
Matched Llama 3.1 405B quality at 70B scale; knowledge cutoff October 2024. Month known, exact day not stated.
Llama 3.2 90B Vision released
First natively multimodal Llama models (11B/90B vision; 1B/3B text-only). Month known, exact day not stated.
Catch Me Up — everything that changed across AI & smart home ↗
Llama 3.2 11B Vision is an open‑weight multimodal model from Meta released in September 2024 as part of the Llama 3.2 family, which also introduced smaller text‑only edge models. It combines 11 billion parameters with a large 128 k token context window and is optimized for strong single‑GPU performance on vision tasks. The model continues Meta’s strategy of providing accessible, high‑capacity AI systems for both language and image processing.
Llama 3.2 90B Vision is an open‑weight model from Meta released in September 2024 as part of the 3.x generation, featuring a 128 000‑token context window and the ability to process images and documents together with text, making it the most capable vision variant in its series under the Llama 3.2 license.
Llama 3.1 8B is an active, fast‑tier model in Meta’s Llama series released on 1 July 2024. It is a small‑parameter variant offering a 128 k token context window and is commonly deployed locally for low‑cost inference through frameworks such as Ollama, vLLM, and llama.cpp.
Llama 3.1 405B is an open‑weight large language model from Meta released in July 2024 and currently active. It expands the Llama 3 series with a 405‑billion‑parameter configuration and supports a context window of up to 128 000 tokens, positioning it as the first openly available model that approaches the performance of proprietary systems such as GPT‑4 when run on high‑end hardware (approximately eight NVIDIA H100 GPUs).
Llama 3.1 70B is a 70‑billion‑parameter language model from Meta released on 1 July 2024 with an extended context window of 128 000 tokens, positioned as a high‑capacity option for many teams before being superseded by Llama 3.3 70B for text tasks. It is now classified as legacy within the Llama family.
Llama 3.2 11B Vision released
Single-GPU vision model from the September 2024 Llama 3.2 release. Month known, exact day not stated.
Llama 3.1 8B released
Part of the July 2024 Llama 3.1 release. Month known, exact day not stated.
Llama 3.1 405B released
Extended context to 128K; 405B first open-weight model genuinely competitive with GPT-4. Month known, exact day not stated.
Llama 3.1 70B released
Part of the July 2024 Llama 3.1 release (8B/70B/405B). Month known, exact day not stated.
Llama 3 released
8B and 70B; major quality jump, 8K context (extended to 128K in 3.1). Month known, exact day not stated.
Llama 2 released
First Llama with commercial-use licensing; sizes 7B/13B/70B, 4K context. Month known, exact day not stated.
LLaMA 1 released
Original Large Language Model Meta AI; research-only weights (7B/13B/33B/65B). Month known, exact day not stated.
