Meta
Llama 4 Maverick
Open SourceMeta's mid-sized open-weights model and the most capable Llama 4 variant for general use. Maverick runs as a mixture-of-experts architecture with 400B total parameters but only 17B active — giving it good speed at 115 t/s while maintaining an AA Intelligence Index of 18. It's multimodal, handles 1M tokens of context, and can be self-hosted. The trade-off: it trails frontier closed models significantly on all AA-measured benchmarks.
Context window
1.0M tokens
API (blended)
$0.31/1M
Consumer access
Free (limited)
Multimodal
Yes
Score Breakdown
40.1/100 → 4.0/10Intelligence, Reliability, Speed, and Context are field-relative — scores shift as models are added. Accessibility and Trust are absolute checklists. Full methodology →
Strengths
- +Open weights (Llama 4 Community License) — auditable, self-hostable, no vendor lock-in
- +1M token context window at a fraction of closed-model pricing
- +Natively multimodal: image + text input in a single API call
- +115 t/s output speed (AA-measured) — fast inference even at large parameter counts
- +Extremely cheap API: ~$0.31/1M blended via Groq
- +Active ecosystem: supported by Groq, Together, Fireworks, Ollama, and others
Weaknesses
- -AA Intelligence Index 18 — significantly trails frontier models (Gemini 3.1 Pro: 57, Claude Opus: 46)
- -GPQA Diamond 67.1%, HLE 4.8% (AA-measured) — limited on hard science and expert reasoning
- -τ²-bench 17.8% (AA-measured) — weak agentic tool use vs frontier alternatives
- -Meta data practices: not ideal for privacy-sensitive enterprise use
- -No MCP support
- -Weights require significant hardware to self-host at full precision
Best for
Not ideal for
Pricing details
Subscription plans
API pricing
Prices verified February 2026. LLM pricing changes frequently — verify at the provider's site before budgeting.
Last updated: February 27, 2026