
LLaMA 4 Scout is a 17-billion parameter model leveraging a Mixture-of-Experts architecture with 16 active experts, positioning it as the top multimodal model in its category. It consistently outperforms competitors like Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1 across diverse benchmark tasks. Despite its performance, LLaMA 4 Scout is remarkably efficient—capable of running on a single NVIDIA H100 GPU with Int4 quantization. It also boasts an industry-leading 10 million token context window and is natively multimodal, seamlessly processing text, images, and video inputs for advanced real-world applications.
| Grok 4 | Llama 4 Scout | |
|---|---|---|
Web Site
| ||
Provider
| ||
Chat
| ||
Release Date
| ||
Modalities
| text images voice video | text images video |
API Providers
| xAI | Meta AI, Hugging Face, Fireworks, Together, DeepInfra |
Knowledge Cut-off Date
| - | 2025-04 |
Open Source
| No | Yes (Source) |
Pricing Input
| $3.00 per million tokens | Not available |
Pricing Output
| $15.00 per million tokens | Not available |
MMLU
| - | Not available |
MMLU-Pro
| - | 74.3% Reasoning & Knowledge Source |
MMMU
| - | 69.4% Image Reasoning Source |
HellaSwag
| - | Not available |
HumanEval
| - | Not available |
MATH
| - | Not available |
GPQA
| 87.5% Science Source | 57.2% Diamond Source |
IFEval
| - | Not available |
SimpleQA
| - | - |
AIME 2024 | - | - |
AIME 2025 | 91.7% Competition Math Source | - |
Aider Polyglot
| - | - |
LiveCodeBench v5
| 79% Competitive Coding Source | - |
Global MMLU (Lite)
| - | - |
MathVista
| - | - |
Mobile Application | - | |
MathArena | ||
| Avg. Score | 89% | - |
AIME 2025 A test based on problems from the American Invitational Mathematics Examination, designed to assess the mathematical skills of models. | 91% | - |
HMMT February 2025 A test based on problems from the Harvard-MIT Mathematics Tournament, February 2025, designed to assess the mathematical skills of models. | 92% | - |
BRUMO 2025 | 95% | - |
SMT 2025 A test based on problems from the Stanford Math Tournament, 2025, designed to assess the mathematical skills of models. | 86% | - |
CMIMC 2025 A test based on problems from the Canadian Mathematical Olympiad, 2025, designed to assess the mathematical skills of models. | 83% | - |
Compare AI. Test. Benchmarks. Mobile Apps Chatbots, Sketch
Copyright © 2026 All Right Reserved.