Llama 4 Scout
Meta · released Apr 5, 2025
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...
Specification
- Context window
- 1.31M
- Max output
- 118K
- Knowledge cutoff
- Aug 31, 2024
- Parameters
- 108.6B
- Licence
- other
- Serving providers
- 3
- Moderated
- No
- Uptime
- 100.0%
Intelligence
6.5
18th percentile
Coding
8.2
Coding Index
Agentic
—
Agentic Index
Output speed
32 t/s
Median across providers
Latency
568ms
Time to first token
Cost per task
$0.02
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- MMLU-Pro75.2%
- GPQA Diamond58.7%
- IFBench39.5%
- LiveCodeBench29.9%
- AA-LCR (long context)27.7%
- SciCode21.3%
- τ²-bench (Telecom)15.5%
- AIME 202514.0%
- Humanity's Last Exam3.8%
- Terminal-Bench Hard1.5%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| MMLU-Pro | 75.2% |
| GPQA Diamond | 58.7% |
| IFBench | 39.5% |
| LiveCodeBench | 29.9% |
| AA-LCR (long context) | 27.7% |
| SciCode | 21.3% |
| τ²-bench (Telecom) | 15.5% |
| AIME 2025 | 14.0% |
| Humanity's Last Exam | 3.8% |
| Terminal-Bench Hard | 1.5% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GLM 4.5VZ.ai6.7
- Gemini 2.5 Flash LiteGoogle6.7
- 6.7
- GPT-4o-miniOpenAI6.7
- GPT-4o-mini (2024-07-18)OpenAI6.7
- Llama 4 ScoutMeta6.5
- SabaMistral AI6.5
- Granite 4.1 8BIBM Granite6.4
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GLM 4.5V | 6.7 |
| Gemini 2.5 Flash Lite | 6.7 |
| Qwen2.5 Coder 32B Instruct | 6.7 |
| GPT-4o-mini | 6.7 |
| GPT-4o-mini (2024-07-18) | 6.7 |
| Llama 4 Scout | 6.5 |
| Saba | 6.5 |
| Granite 4.1 8B | 6.4 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.18
- Output / 1M tokens
- $0.59
- Cached input / 1M
- Not offered
- Blended 3:1
- $0.282
One task, estimated
$0.02
- Input tokens
- 50,000
- Output tokens
- 25,000
- Profile
- Standard
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 776
- Win rate
- 27.4%
- Strongest at
- Game development
- Tournaments
- 175
Elo by category
Dot position on a 750–800 scale · Elo has no meaningful zero
- 776
- 769
- 752
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| Game development | 776 | 27.4% |
| UI components | 769 | 25.5% |
| Websites | 752 | 22.7% |