Early build · scores computed by the deterministic engine from real, sourced developments (30 days) · weights v1.0-prior
Live
☰
Latest › Models › Meta
Vikshy Models · Meta

Meta models: pricing, context & what each is for

Meta has 8 models in the Vikshy catalogue, spanning Llama 4, Llama 3, compared side by side on context window, API pricing, modality, and what each is best for, every model linked to Meta's official docs and pricing. Prices are per million tokens and can change, so the linked pricing page is the source of truth. For how Meta ranks against other labs, see its Vikshy Score. Updated 2026-07-21.

ModelContextInput /MOutput /MModalityAccess
Llama 4 MaverickFlagship
General assistant and multimodal workloads run through third-party hosts or self-hosting.
1M
Open
Open
Multimodal
Open weights + third-party API
DocsPrice
Llama 4 ScoutEfficient
Very long context tasks and efficient on device or single GPU deployment.
10M
Open
Open
Multimodal
Open weights + third-party API
DocsPrice
Llama 3.3 70BLegacy
Teams wanting a proven mid size open text model.
128K
Open
Open
Text
Open weights + third-party API
DocsPrice
Llama 3.2 90B VisionLegacy
Multimodal apps that need self hosted image plus text.
128K
Open
Open
Vision
Open weights + third-party API
DocsPrice
Llama 3.2 11B VisionLegacy
Lighter multimodal workloads and edge image tasks.
128K
Open
Open
Vision
Open weights + third-party API
DocsPrice
Llama 3.2 3BLegacy
Edge and mobile assistants where footprint matters.
128K
Open
Open
Text
Open weights + third-party API
DocsPrice
Llama 3.1 405BLegacy
Research and high end self hosted deployments.
128K
Open
Open
Text
Open weights + third-party API
DocsPrice
Llama 3.1 8BLegacy
Cost sensitive text tasks and fine tuning starters.
128K
Open
Open
Text
Open weights + third-party API
DocsPrice
Meta's Vikshy Score →All models, compared →What is the Vikshy Score? →

Meta models, common questions

What is Meta's most capable model?

Its highest-tier model in the catalogue is Llama 4 Maverick (Flagship tier), built for general assistant and multimodal workloads run through third-party hosts or self-hosting.

What is the largest context window Meta offers?

Llama 4 Scout has the largest listed context window, up to 10M tokens.

Does Meta offer open-weight models?

Yes. 8 of Meta's catalogued models are available as open weights.