Model intelligence
Meta's Llama 4 Scout — 17B active / 109B total Mixture-of-Experts (16 experts) with a 10M-token context window. Natively multimodal: text and up to 5 images as input; text-only output. Open-weight model, not a hosted paid API.
Data freshness: last checked 8 d ago · updated 2026-05-21
5 of 10 tracked fields verified against a primary source.
Data freshness: last checked 8 d ago · updated 2026-05-21
Verification checklist
5 of 10 tracked fields verified against a primary source
Each row reflects the actual state on the record — no checkmark is asserted without isVerified() returning true.
API identifiers (canonical / alias)
Context window
Max output tokens
Modality channels
Knowledge cutoff
Features (extended thinking / tool use / vision …)
Lifecycle (active / deprecated / retired)
Release date
Snapshot date
Pricing rows
0 pricing rows sourced
Data gaps
These fields are intentionally null until verified against a primary source. They are not estimated.
Max output tokens
Provider docs do not publish a max-output number separately from the context window for this model, or the field has not yet been retrieved.
Pricing
No pricing rows verified yet. Provider pricing pages may render with JavaScript only, or may not publish per-model rates at a stable URL.
Benchmark scores
No third-party benchmark scores have been verified for any model on this site — see the benchmark-limitations guide for why.
Infrastructure (regions / latency / uptime)
Infrastructure fields are intentionally null until a durable measurement source is wired. See the inference-infrastructure guide.
How a value moves from this list to a verified row — see source verification methodology.
Reference
Pinned snapshot IDs as listed by the provider.
API pricing
Per-unit rates published by Meta for its own first-party API. Each row is a source-backed reference, not a live quote. Where Meta does not run a paid first-party API for this model, the table renders the unverified-data label and any hosted-provider rates appear below.
Pricing is volatile and may change after retrieval. Treat prices as source-backed references, not live quotes.
Data not yet verified.Third-party platforms that host this model and bill at their own rate. Hosted pricing is billing-provider-specific and may change. It is not model-creator pricing unless the creator is also the billing provider. See methodology.
Hosted by Groq
meta-llama/llama-4-scout-17b-16e-instructGroq advertises additional features alongside the per-token rate (50% cache-hit discount on uncached input tokens, 50%-off Batch API with a 24h–7d window). Those are documented on the pricing page but the catalogue keeps each platform's promo modifiers in record notes rather than synthesising derived rows.
Capability
Benchmark scores are only published once verified against a primary source. WebmasterID Models does not republish provider-reported scores without an independent reference.
Infrastructure
Regions, latency, and uptime where independently verified.
Regions
Data not yet verified.
Avg latency (ms)
Data not yet verified.
Uptime (%)
Data not yet verified.
Continue
Other models in the catalogue — same provider first, then the rest of the graph.
Entity graph
Meta Llama 4 model card (Scout + Maverick). Context windows (Scout 10M, Maverick 1M), input modalities (text + up to 5 images), output modality (text only), parameter counts (17B active), knowledge cutoff (August 2024). Max output tokens and a stated release date are NOT on the page. Meta does not run a hosted paid API for Llama; pricing stays null.
Next review for this entity targets re-checking pricing against official provider documentation and confirming lifecycle state. Cadence per field type lives in /docs.