Llama 4 Scout
Meta AIhuggingface.co/meta-llama/Llama-4-Scout-17B-16E-
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...
Updated 15 min ago · first seen 11 Sept 2026
model_01M294WWHQ4KJCH5S61NESVPX9
- Parameters
- 108.6B
- T2 · 2 h ago
- Context
- 10M tokens
- T2 · 26 min ago
- Released
- 5 Apr 2025
- T2 · 30 min ago
- License
- other
- T2 · 1 h ago
- Knowledge cutoff
- Aug 2024
- T2 · 2 h ago
Specification
- Release date
- 5 Apr 2025
Source:OpenRouter public model & pricing listingT2observed 30 min agomedium
- Openness
- restricted
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 min agomedium
- License
- other
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Architecture
- Llama4ForConditionalGeneration
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Parameters
- 108.6B
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Context window
- 10M tokens
Source:Artificial AnalysisT2observed 26 min agomedium
- Max output
- 16.4K tokens
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Knowledge cutoff
- Aug 2024
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Modalities
- image, text
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Input modalities
- image, text
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Output modalities
- text
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Languages
- ar, de, en, es, fr, hi, id, it, pt, th, tl, vi
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Tokenizer
- Llama4
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Base model
- meta-llama/Llama-4-Scout-17B-16E
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- File size
- 217.3 GB
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Hugging Face repo
- meta-llama/Llama-4-Scout-17B-16E-Instruct
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Pipeline tag
- image-text-to-text
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Downloads
- 161,484
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 min agomedium
- Likes
- 1,345
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 min agomedium
- Gated
- manual
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Hf inference providers
- deepinfra, groq, novita, nscale
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Last modified
- 2025-05-22T23:44:50+00:00
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Library name
- transformers
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Aa median output tokens per second
- 95.5
Source:Artificial AnalysisT2observed 26 min agomedium
- Downloads all time
- 7,621,418
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Model type
- llama4
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Openrouter id
- meta-llama/llama-4-scout
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Supported parameters
- frequency_penalty, logit_bias, max_tokens, min_p, presence_penalty, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_p
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Tags
- transformers, safetensors, llama4, image-text-to-text, facebook, meta, pytorch, llama, ar, de, en, es, fr, hi, id, it, pt, th, tl, vi
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Weights dtype
- BF16
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →
Provenance
Attributed facts
36
Source tiers
T236
Freshest observation
15 min ago
Conflicts
None
Modalities
- Modalities
- imagetext
- Input
- imagetext
- Output
- text
Capabilities
Tool calling
Yes
OpenRouter public model & pricing listing · T2
Structured output
Yes
OpenRouter public model & pricing listing · T2
Reasoning
No
Artificial Analysis · T2
Vision
Yes
OpenRouter public model & pricing listing · T2
Audio
Unavailable
Fine-tuning available
Unavailable
- Context window
- 10M tokens
Source:Artificial AnalysisT2observed 26 min agomedium
- Max output
- 16.4K tokens
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Knowledge cutoff
- Aug 2024
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Languages
- ar, de, en, es, fr, hi, id, it, pt, th, tl, vi
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 1 h agomedium
- Tokenizer
- Llama4
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
| Benchmark | Score | Metric | Config | Evaluated | Source |
|---|---|---|---|---|---|
| τ²-bench | 15.5% | pass^1 | aa_slug=llama-4-scout · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Artificial Analysis Intelligence Index | 6.45 | index | aa_slug=llama-4-scout · version=4.3 · estimated=false | — | artificialanalysis.aiT2 |
| GPQA | 58.69% | accuracy | aa_slug=llama-4-scout · variant=GPQA Diamond · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Humanity's Last Exam | 3.78% | accuracy | aa_slug=llama-4-scout · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| IFBench | 39.52% | accuracy | aa_slug=llama-4-scout · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| MMMU-Pro | 52.95% | accuracy | aa_slug=llama-4-scout · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| SciCode | 21.3% | accuracy | aa_slug=llama-4-scout · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Terminal-Bench | 0% | accuracy | aa_slug=llama-4-scout · variant=v4.0 · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Terminal-Bench | 3.75% | accuracy | aa_slug=llama-4-scout · variant=v2.1 · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Terminal-Bench | 1.52% | accuracy | aa_slug=llama-4-scout · variant=hard · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
Scores are reported as published, with their evaluation configuration (harness, prompting, judge). Results with different configs are not directly comparable — see methodology.
Current prices
| Provider | Input / 1M | Output / 1M | Cached in | Batch in / out | Context | Observed | Source |
|---|---|---|---|---|---|---|---|
| OpenRoutermeta-llama/llama-4-scout | $0.10 | $0.30 | — | — | 327.7K | 30 min ago | openrouter.aiT2 |
USD per 1M tokens as published by each provider (USD). Rows are append-only: every change is kept in the history below.
Available through 1
Price history
Input price · USD / 1M tokens
Output price · USD / 1M tokens
- OpenRouter
Memory need = bytes per parameter (4-bit ≈ 0.5 × 1.15 overhead, 8-bit 1.0, fp16 2.0) + a KV-cache allowance. Not a measurement.
| Hardware | Quantization | Memory | Est. need | Fits |
|---|---|---|---|---|
| NVIDIA DGX B200 | 4bit | 1,440 GB | 63.0 GB | Yes |
| Apple M3 Ultra | 4bit | — | 63.0 GB | Yes |
| Mac Studio (Apple M5 Ultra) | 4bit | — | 63.0 GB | Yes |
| AMD Instinct MI325X | 4bit | 256 GB | 63.0 GB | Yes |
| AMD Instinct MI300X | 4bit | 192 GB | 63.0 GB | Yes |
| Apple M2 Ultra | 4bit | — | 63.0 GB | Yes |
| NVIDIA B200 | 4bit | 180 GB | 63.0 GB | Yes |
| NVIDIA H200 | 4bit | 141 GB | 63.0 GB | Yes |
| NVIDIA H200 NVL | 4bit | 141 GB | 63.0 GB | Yes |
| Apple M1 Ultra | 4bit | — | 63.0 GB | Yes |
| Apple M3 Max | 4bit | — | 63.0 GB | Yes |
| Apple M4 Max | 4bit | — | 63.0 GB | Yes |
| Mac Studio (Apple M5 Max) | 4bit | — | 63.0 GB | Yes |
| MacBook Pro (Apple M5 Max) | 4bit | — | 63.0 GB | Yes |
| NVIDIA DGX Spark | 4bit | 128 GB | 63.0 GB | Yes |
| Apple M2 Max | 4bit | — | 63.0 GB | Yes |
| NVIDIA H100 NVL | 4bit | 94 GB | 63.0 GB | Yes |
| NVIDIA A100 80GB | 4bit | 80 GB | 63.0 GB | Yes |
| NVIDIA H100 SXM | 4bit | 80 GB | 63.0 GB | Yes |
| Apple M1 Max | 4bit | — | 63.0 GB | No |
| Apple M4 Pro | 4bit | — | 63.0 GB | No |
| Mac mini (Apple M5 Pro) | 4bit | — | 63.0 GB | No |
| MacBook Pro (Apple M5 Pro) | 4bit | — | 63.0 GB | No |
| Apple M3 Pro | 4bit | — | 63.0 GB | No |
| Apple M1 Pro | 4bit | — | 63.0 GB | No |
| Apple M2 Pro | 4bit | — | 63.0 GB | No |
| Apple M4 | 4bit | — | 63.0 GB | No |
| iMac (Apple M4) | 4bit | — | 63.0 GB | No |
| Mac mini (Apple M6) | 4bit | — | 63.0 GB | No |
| MacBook Air (Apple M5) | 4bit | — | 63.0 GB | No |
| MacBook Pro (Apple M5) | 4bit | — | 63.0 GB | No |
| NVIDIA GeForce RTX 5090 | 4bit | 32 GB | 63.0 GB | No |
| Apple M2 | 4bit | — | 63.0 GB | No |
| Apple M3 | 4bit | — | 63.0 GB | No |
| NVIDIA GeForce RTX 3090 | 4bit | 24 GB | 63.0 GB | No |
| NVIDIA GeForce RTX 4090 | 4bit | 24 GB | 63.0 GB | No |
| Apple M1 | 4bit | — | 63.0 GB | No |
Ancestors 1
- Llama-4-Scout-17B-16EMeta AI
This model
Llama 4 Scout
108.6B params
Quantizations 0
None recorded.
Descendants 0
None recorded.
Papers 1
- arXiv:2204.05149Active35
Repositories 0
No repositories linked yet.
As of
Rewind the record: see this entity's attributes exactly as AI Atlas knew them on a given day.
Claim history · Model card
Model cardmodel_card_url1
| Value | Valid from → to | Status | Source | Confidence | Extractor |
|---|---|---|---|---|---|
| https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct | → current | current | Hugging Face Hub (public pages, model cards, papers)T2 | medium | deterministic |
Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →
Llama 4 Scout: openness changed from open-weights to restricted
Opennessopen-weights→restrictedhuggingfaceLlama 4 Scout scores 1.52% on Terminal-Bench
artificial_analysisLlama 4 Scout scores 3.75% on Terminal-Bench
artificial_analysisLlama 4 Scout scores 0% on Terminal-Bench
artificial_analysisLlama 4 Scout scores 3.78% on Humanity's Last Exam
artificial_analysisLlama 4 Scout scores 6.45 on Artificial Analysis Intelligence Index
artificial_analysisLlama 4 Scout: context length changed from 1310720 to 10000000
Context window1.31M tokens→10M tokensartificial_analysisLlama 4 Scout: openness changed from restricted to open-weights
Opennessrestricted→open-weightsartificial_analysisLlama 4 Scout: release date changed from 2025-04-02 to 2025-04-05
Release date2 Apr 2025→5 Apr 2025openrouterLlama 4 Scout: release date changed from 2025-04-05 to 2025-04-02
Release date5 Apr 2025→2 Apr 2025huggingfaceOpenRouter lists Llama 4 Scout at $0.1 in / $0.3 out per 1M tokens
openrouter
| Source | Document | Type | Tier | Last observed | Snapshots |
|---|---|---|---|---|---|
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/models?author=meta-llama&p=0&sort=downloads | listing | T2· Quality secondary | 15 min ago | 3 |
| Artificial Analysis | artificialanalysis.ai/leaderboards/models | leaderboard | T2· Quality secondary | 26 min ago | 1 |
| OpenRouter public model & pricing listing | openrouter.ai/api/v1/models | catalogue | T2· Quality secondary | 30 min ago | 2 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct | model_page | T2· Quality secondary | 1 h ago | 1 |
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.