Skip to content

status.openkey.ai is live — independent, worldwide uptime proof See it now ->

openkey

Meta AI

12 models built

Meta (Meta Platforms, Inc., founded 2004, based in the United States) builds the Llama family of open-weight models, available through OpenKey under the meta-llama namespace. The lineup spans four generations, from Llama 3 through Llama 4, plus a dedicated moderation model, Llama Guard 4 12B. Meta's strategy is breadth over a single flagship: small models like Llama 3.2 1B for cheap, fast inference, mid-size 70B-class models for general use, and Llama 4 Maverick and Scout for long-context, multimodal work — Scout supports a 10M-token context window, the largest in the roster. Two models, Llama 3.2 3B Instruct and Llama 3.3 70B Instruct, are offered free. OpenKey lists 12 Meta models total.

Lineup

For general text tasks at low cost, Llama 3.1 8B Instruct is the cheapest paid option at $0.02/$0.03 per 1M input/output tokens (OpenKey: $0.0206/$0.0309 after the 3% fee). If budget is the only constraint, use the free tiers: Llama 3.3 70B Instruct (free) or Llama 3.2 3B Instruct (free) cost nothing. For vision or long-context work, Llama 4 Scout ($0.10/$0.30 per 1M, OpenKey $0.103/$0.309) handles up to 10M tokens of context, while Llama 4 Maverick ($0.15/$0.60 per 1M, OpenKey $0.1545/$0.618) trades some context (1,048,576 tokens) for stronger multimodal output. For content moderation, use Llama Guard 4 12B at $0.18/$0.18 per 1M (OpenKey $0.1854/$0.1854).

Prices per 1M tokens, flat 3% fee included.

Questions

What's the price range across Meta's models on OpenKey?
Paid Meta models range from Llama 3.1 8B Instruct at $0.02/$0.03 per 1M input/output tokens up to Llama 3.1 70B Instruct at $0.40/$0.40 per 1M. On OpenKey, that's provider price × 1.03, so Llama 3.1 8B Instruct comes to $0.0206/$0.0309 and Llama 3.1 70B Instruct to $0.412/$0.412 per 1M tokens. Two models, Llama 3.3 70B Instruct and Llama 3.2 3B Instruct, are free.
Which Meta model should I use if I want the cheapest paid option?
Llama 3.1 8B Instruct is the cheapest paid Meta model at $0.02 per 1M input tokens and $0.03 per 1M output tokens — $0.0206/$0.0309 on OpenKey after the 3% fee. If you don't need paid-tier guarantees, Llama 3.3 70B Instruct (free) or Llama 3.2 3B Instruct (free) cost nothing at all.
Is there a Meta model built for long-context or moderation tasks?
Llama 4 Scout supports up to 10,000,000 tokens of context at $0.10/$0.30 per 1M input/output ($0.103/$0.309 on OpenKey). Llama 4 Maverick handles 1,048,576 tokens at $0.15/$0.60 per 1M ($0.1545/$0.618 on OpenKey). For moderation specifically, Llama Guard 4 12B is priced at $0.18/$0.18 per 1M ($0.1854/$0.1854 on OpenKey) and supports text+image input.