Llama 4: replaced zero token prices with null: downloadable weights do not establish free hosted inference or a universal API price

The OnAirToday AI Progress Tracker recorded a verified change to Meta's Llama 4 (version Scout / Maverick): Replaced zero token prices with null: downloadable weights do not establish free hosted inference or a universal API price.

Llama 4 Scout and Maverick are Meta's downloadable-weight, multimodal mixture-of-experts models; the benchmarks below are Meta's published Maverick release results, with AIME referring to AIME 2024. Live source retrieval was unavailable for this response, so subsequent releases and the supplied 2026 incident reports could not be independently verified as of 20 September 2026.

What changed in this review:

• Identified stored benchmark figures as Maverick release results; AIME refers to AIME 2024.
• Recorded supplied security disclosures and copyright allegations with attribution, duplicate-coverage warnings, and no claim of confirmed in-the-wild exploitation.
• Excluded the aggregate incident-count headline because it identifies no specific additional case.
• Current release status and supplied 2026 reports remain unverified without live source access.
• Flaws and limitations updated
• Risks and threats updated

Teams building on Llama 4 should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.

The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/llama-4.

References:

• https://ai.meta.com/blog/llama-4-multimodal-intelligence/
• https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
• https://huggingface.co/meta-llama/Llama-4-Maverick-17B-128E-Instruct