HOST DIAGNOSTIC · Meta
Llama 4 Scout
Llama 4 Scout· Llama· OPEN WEIGHTS
ACTIVE
HOST IDENTIFICATION
Meta
Llama 4 Scout
ACTIVE
AVG SCORE
[11.0]
CAPABLE
STRONGEST
Cost Efficiency
[16]
WEAKEST
Self-Correction
[9]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
11
Reasoning
11
Mathematical Precision
10
World Knowledge
11
Scientific Acumen
10
Technical
Code Generation
11
Tool Use
10
Multimodal Fluency
11
Speed
15
Cost Efficiency
16
Behavioral
Candor
10
Creativity
10
Tenacity
9
Self-Correction
9
Calibration
10
Operational
Instruction Following
11
Context Fidelity
15
Multi-turn Coherence
10
Planning and Execution
10
Safety Alignment
10
SCORING METHODOLOGY
RECALIBRATED. Gen 1 scored Scout HIGHER than Maverick (15-16 vs 7-8) which is backwards — Scout is the smaller, efficiency-focused variant. Recalibrated to 11-12 range. Context Fidelity kept high at 16 (10M token window is genuinely exceptional). Cost 18 + Speed 16 validated (efficiency is Scout's value prop). Open weights, part of Llama 4 family.