At a Glance
| Vendor | Meta |
| Family | Llama 4 series |
| Launched | Meta released the Llama 4.5 family in early June 2026 as a mid-cycle refresh of the Llama 4 Scout and Maverick variants. The update brings incremental capability gains, improved agentic-tool-use stability, and modest license clarifications without changing the broad community-license terms. |
| Context window | Scout 4.5: 10,000,000 tokens maintained. Maverick 4.5: 1,000,000 tokens, with a long-context variant under evaluation. |
| Pricing | Free to download and self-host under Meta's community license. Commercial use restrictions at very large scale unchanged. Hosted API pricing via Groq, Together AI, Fireworks, Vertex AI, AWS Bedrock remains aggressive. |
| Access channels | Direct download (Hugging Face, llama.com), Meta AI consumer products (WhatsApp, Instagram, Messenger, Ray-Ban Meta, Threads), Groq, Together AI, Vertex AI, AWS Bedrock, Azure AI, and essentially every major cloud and inference provider. |
Notable Benchmarks
Incremental gains on general-capability benchmarks. The headline improvements are on agentic-tool-use stability and reduced hallucination on long-context retrieval, addressing the most-cited Llama 4 production issues.
Strengths
Open weights, extremely wide deployment surface across Meta consumer products and cloud partners, large context windows enabling long-document and multi-file workflows, improved agentic stability over Llama 4.0.
Limitations
Benchmark gaps versus frontier closed-models persist on the most complex reasoning. Community license restrictions on commercial scale occasionally matter for the very largest deployers. Llama 4.5 is incremental, not a step-change.
Brand-Visibility Implications
Llama 4.5 powers Meta AI across WhatsApp, Instagram, Messenger, and Threads, a combined audience of several billion users. For any global B2C brand, Llama 4.5 visibility via Meta AI is material consumer-discovery surface. For brands building on open-weight infrastructure, Llama 4.5 is the default choice for most use cases where Apache 2.0 licensing (Gemma 4) or permissive Chinese licensing (Qwen 3.7, DeepSeek V4.1) is not strictly required. See Llama visibility and Llama 4 brief.
How Presenc AI Tracks This Model
Presenc AI monitors brand visibility on Meta's Llama 4 series as part of continuous multi-platform AI visibility tracking. We sample Meta Llama 4.5 across representative prompt sets daily, compare against competitor performance on the same prompts, and flag material mention-rate changes so brand teams can respond quickly when AI representation shifts.