About
Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B. It supports native multimodal input...
Related Models
Pricing
Input
$0.10
per 1M tokens
Output
$0.30
per 1M tokens
Blended
$0.15
per 1M tokens
Cheaper than 75% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$0.15
Monthly
$4.50
vs. Similar Models
Qwen: Qwen3 VL 30B A3B InstructQ:0.0
$0.23+52%
Hermes 4 - Llama-3.1 70B (Reasoning)Q:0.0
$0.20+32%
Qwen3 14B (Reasoning)Q:+0.1
$0.73+387%
Claude 3.5 Sonnet (Oct '24)Q:-0.1
$6.00+3900%
Performance
109
tokens/sec
Faster than 58% of models
0.61
seconds
Faster than 76% of models
0.61
seconds
Faster than 83% of models
Market Median
94 tok/s
16% faster
Median TTFT
1.11s
45% faster
Throughput/Dollar
730
tok/s per $/1M
Speed Comparison
Llama 3.2 Instruct 11B (Vision)
110 tok/s+0%
GPT-5.4 (low)
109 tok/s-0%
Kwaipilot: KAT-Coder-Pro V2
109 tok/s-0%
Context Window
10.0M
tokens
Larger than 100% of models
Max Output
16K
tokens
0% of context
Benchmarks
MMLU-Pro
75.2%
GPQA Diamond
58.7%
HLE
4.3%
LiveCodeBench
29.9%
SciCode
17.0%
TerminalBench Hard
1.5%
MATH-500
84.4%
AIME
28.3%
AIME 2025
14.0%
IFBench
39.5%
Long Context Recall
25.8%
Tau2
15.5%
Market AverageTop Score