Meta API Pricing

Llama 3.2 1B is one of the smallest capable LLMs — sub-$0.05 per million tokens and runs on CPUs. Useful for classification and routing at extreme scale or very constrained hardware.

Llama 3.1 8B is the smallest open-weight Llama 3.1 model — extremely cheap to host or call, and good enough for classification, extraction, and basic chat.

Llama 3.2 3B is a tiny but surprisingly capable open-weight model — one of the cheapest LLMs available from any provider. Fits on edge hardware and consumer GPUs with room to spare.

Llama 4 Scout is Meta's latest MoE model with an industry-leading 10M token context window at an affordable price. Remarkable context-to-cost ratio — suitable for entire-codebase and very long document tasks.

Llama 3.3 70B improves on Llama 3.1 70B with better instruction-following and reasoning, at the same price point. The recommended Llama 70B for new projects — same hosting cost, meaningfully better quality.

Llama 3.2 11B Vision is Meta's small open-weight multimodal model — capable of understanding images at a fraction of GPT-4o's cost. The go-to for budget image + text pipelines.

Llama Guard 4 is a Llama 4 Scout-derived multimodal pretrained model, fine-tuned for content safety classification.

Llama 4 Maverick is Meta's flagship open-weight model in the Llama 4 generation — multimodal, 1M context, and competitive with GPT-4o at a fraction of the API cost.

Llama 3.1 70B is the mid-size open-weight model in the 3.1 family — a popular sweet spot for production workloads that need GPT-4o-mini-class quality at open-weight prices.

Llama 3.2 90B Vision is Meta's large open-weight multimodal model — strong on image understanding and visual reasoning while remaining self-hostable. Best open multimodal option before Llama 4.

Llama 3.1 405B is Meta's largest open-weight model — competitive with GPT-4-class models on many benchmarks and uniquely available for self-hosting. Symmetric input/output pricing is common across hosted providers.