Meta: Llama 4 Scout

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 (text and…

meta-llama/llama-4-scout-17b-16e-instruct

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 (text and image) and multilingual output (text and code) across 12 supported languages. Designed for assistant-style interaction and visual reasoning, Scout uses 16 experts per forward pass and features a context length of 10 million tokens, with a training corpus of ~40 trillion tokens.

Built for high efficiency and local or commercial deployment, Llama 4 Scout incorporates early fusion for seamless modality integration. It is instruction-tuned for use in multilingual chat, captioning, and image understanding tasks. Released under the Llama 4 Community License, it was last trained on data up to August 2024 and launched publicly on April 5, 2025.

Specifications & capabilities

Input
text, image
Output
text
Context window · tokens
1310720
Knowledge cutoff
2024-08-31
Published model ID
Llama-4-Scout-17B-16E-Instruct
Capabilities
multimodal, structuredOutputs, toolCalling, visionInput
Description source ↗

Evaluations

1–20 / 95

LLM Stats · GitHub archive

0.944

source id
llmStatsArchive
leaderboard name
DocVQA · Raw
dimension
general
metric
DocVQA · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
DocVQA
language
en
modality
multimodal
verified
false
max_score
1
categories
  • vision
  • multimodal
created_at
2025-07-19T19:56:12.825214+00:00
paper_link
https://arxiv.org/abs/2007.00398
updated_at
2025-07-19T19:56:12.825214+00:00
description
A dataset for Visual Question Answering on document images containing 50,000 questions defined on 12,000+ document images. The benchmark tests AI's ability to understand document structure and content, requiring models to comprehend document layout and perform information retrieval to answer questions about document images.
benchmark_id
docvqa
multilingual
false
analysis_method
0-shot (ANLS)
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
DocVQA · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://ai.meta.com/blog/llama-4-multimodal-intelligence/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.944

source id
llmStatsArchive
leaderboard name
DocVQA · Normalized (reported)
dimension
general
metric
DocVQA · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
DocVQA
language
en
modality
multimodal
verified
false
max_score
1
categories
  • vision
  • multimodal
created_at
2025-07-19T19:56:12.825214+00:00
paper_link
https://arxiv.org/abs/2007.00398
updated_at
2025-07-19T19:56:12.825214+00:00
description
A dataset for Visual Question Answering on document images containing 50,000 questions defined on 12,000+ document images. The benchmark tests AI's ability to understand document structure and content, requiring models to comprehend document layout and perform information retrieval to answer questions about document images.
benchmark_id
docvqa
multilingual
false
analysis_method
0-shot (ANLS)
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
DocVQA · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://ai.meta.com/blog/llama-4-multimodal-intelligence/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.572

source id
llmStatsArchive
leaderboard name
GPQA · Raw
dimension
general
metric
GPQA · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
GPQA
language
en
modality
text
verified
false
max_score
1
categories
  • reasoning
  • general
created_at
2025-07-19T19:56:11.588605+00:00
paper_link
https://arxiv.org/abs/2311.12022
updated_at
2025-07-19T19:56:11.588605+00:00
description
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
benchmark_id
gpqa
multilingual
false
analysis_method
0-shot (accuracy)
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
GPQA · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.572

source id
llmStatsArchive
leaderboard name
GPQA · Normalized (reported)
dimension
general
metric
GPQA · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
GPQA
language
en
modality
text
verified
false
max_score
1
categories
  • reasoning
  • general
created_at
2025-07-19T19:56:11.588605+00:00
paper_link
https://arxiv.org/abs/2311.12022
updated_at
2025-07-19T19:56:11.588605+00:00
description
A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.
benchmark_id
gpqa
multilingual
false
analysis_method
0-shot (accuracy)
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
GPQA · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.678

source id
llmStatsArchive
leaderboard name
MBPP · Raw
dimension
general
metric
MBPP · Raw
rank
2
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
MBPP
language
en
modality
text
verified
false
max_score
100
categories
  • reasoning
  • general
created_at
2025-07-19T19:56:13.453370+00:00
paper_link
https://arxiv.org/abs/2108.07732
updated_at
2025-07-19T19:56:13.453370+00:00
description
MBPP (Mostly Basic Python Problems) is a benchmark of 974 crowd-sourced Python programming problems designed to be solvable by entry-level programmers. Each problem consists of a task description, code solution, and 3 automated test cases covering programming fundamentals and standard library functionality.
benchmark_id
mbpp
multilingual
false
analysis_method
3-shot pass@1
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
MBPP · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
4
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.678

source id
llmStatsArchive
leaderboard name
MBPP · Normalized (reported)
dimension
general
metric
MBPP · Normalized (reported)
rank
2
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
MBPP
language
en
modality
text
verified
false
max_score
100
categories
  • reasoning
  • general
created_at
2025-07-19T19:56:13.453370+00:00
paper_link
https://arxiv.org/abs/2108.07732
updated_at
2025-07-19T19:56:13.453370+00:00
description
MBPP (Mostly Basic Python Problems) is a benchmark of 974 crowd-sourced Python programming problems designed to be solvable by entry-level programmers. Each problem consists of a task description, code solution, and 3 automated test cases covering programming fundamentals and standard library functionality.
benchmark_id
mbpp
multilingual
false
analysis_method
3-shot pass@1
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
MBPP · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
4
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.315

source id
llmStatsArchive
leaderboard name
TydiQA · Normalized (reported)
dimension
general
metric
TydiQA · Normalized (reported)
rank
2
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
TydiQA
language
en
modality
text
verified
false
max_score
1
categories
  • language
  • reasoning
created_at
2025-07-19T19:56:14.470500+00:00
paper_link
https://arxiv.org/abs/2003.05002
updated_at
2025-07-19T19:56:14.470500+00:00
description
A multilingual question answering benchmark covering 11 typologically diverse languages with 204K question-answer pairs. Questions are written by people seeking genuine information and data is collected directly in each language without translation to test model generalization across diverse linguistic structures.
benchmark_id
tydiqa
multilingual
true
analysis_method
1-shot average/f1
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
TydiQA · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
2
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

LLM Stats · GitHub archive

0.315

source id
llmStatsArchive
leaderboard name
TydiQA · Raw
dimension
general
metric
TydiQA · Raw
rank
2
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
TydiQA
language
en
modality
text
verified
false
max_score
1
categories
  • language
  • reasoning
created_at
2025-07-19T19:56:14.470500+00:00
paper_link
https://arxiv.org/abs/2003.05002
updated_at
2025-07-19T19:56:14.470500+00:00
description
A multilingual question answering benchmark covering 11 typologically diverse languages with 204K question-answer pairs. Questions are written by people seeking genuine information and data is collected directly in each language without translation to test model generalization across diverse linguistic structures.
benchmark_id
tydiqa
multilingual
true
analysis_method
1-shot average/f1
is_self_reported
true
verified_by_llmstats
false
attribution
Jonathan Chavez Tamales and LLM Stats contributors (copyright 2024 jc); TensorTrace structures original results and derives scoped ranks.
license_url
https://creativecommons.org/licenses/by/4.0/
metric_kind
TydiQA · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-4-scout/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-4-Scout-17B-16E-Instruct
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
meta/llama-4-scout
metric_population
2
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-4-scout
Original result ↗

BFCL

100%

source id
bfcl
leaderboard name
BFCL Relevance Detection
dimension
toolUse
metric
Relevance detection
rank
6
Original result ↗

BFCL

89.38%

source id
bfcl
leaderboard name
BFCL Non-Live AST
dimension
toolUse
metric
Non-live AST accuracy
rank
8
Original result ↗

LMArena

1,137.931

source id
lmarena
leaderboard name
LMArena Vision Style Control Creative Writing
dimension
multimodal
metric
Arena rating
rank
31
sample size
265
evaluated at
2026-01-09T00:00:00.000Z
source model id
llama-4-scout-17b-16e-instruct
confidence lower
1101.1497480538835
confidence upper
1174.7126181478243
Original result ↗

LMArena

1,126.76

source id
lmarena
leaderboard name
LMArena Vision Creative Writing
dimension
multimodal
metric
Arena rating
rank
32
sample size
265
evaluated at
2026-01-09T00:00:00.000Z
source model id
llama-4-scout-17b-16e-instruct
confidence lower
1090.1427705507765
confidence upper
1163.3765926469855
Original result ↗

BFCL

14.5%

source id
bfcl
leaderboard name
BFCL Web Search
dimension
agentic
metric
Web-search accuracy
rank
49
Original result ↗

BFCL

9%

source id
bfcl
leaderboard name
BFCL Multi Turn
dimension
agentic
metric
Multi-turn accuracy
rank
71
Original result ↗

LMArena

1,100.572

source id
lmarena
leaderboard name
LMArena Vision Style Control Humor
dimension
multimodal
metric
Arena rating
rank
100
sample size
191
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 4 Scout 17B 16E Instruct
confidence lower
1055.6987914818444
confidence upper
1145.444234737787
Original result ↗

LMArena

1,100.801

source id
lmarena
leaderboard name
LMArena Vision Humor
dimension
multimodal
metric
Arena rating
rank
101
sample size
191
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 4 Scout 17B 16E Instruct
confidence lower
1055.789016334109
confidence upper
1145.8133628038552
Original result ↗

BFCL

44.92%

source id
bfcl
leaderboard name
BFCL Irrelevance Detection
dimension
toolUse
metric
Irrelevance detection
rank
101
Original result ↗

Published prices

1–6 / 6

Provider specifications & access

1–3 / 3

llama-4-scout-17b-16e-instruct ↗

source model id
@cf/meta/llama-4-scout-17b-16e-instruct
hosting
cloudflare-hosted
platforms
  • Cloudflare Workers AI
context length
131000
architecture
input_modalities
  • text
  • image
output_modalities
  • text
capabilities
  • toolCalling

Llama-4-Scout-17B-16E-Instruct ↗

source model id
publishers/hf-meta-llama/models/llama-4-scout-17b-16e-instruct@001
hosting
vertex-self-deployed
platforms
  • Google Cloud Vertex AI
architecture
input_modalities
  • text
output_modalities
  • text
capabilities

    Llama-4-Scout-17B-16E-Instruct ↗

    source model id
    publishers/meta/models/llama4@llama-4-scout-17b-16e-instruct
    hosting
    vertex-self-deployed
    platforms
    • Google Cloud Vertex AI
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities

      Safety assessments

      1–1 / 1

      Foundation Model Transparency Index 2025 ↗

      31 / 100

      source id
      stanfordFMTI
      canonical model id
      meta-llama/llama-4-scout-17b-16e-instruct
      taxonomy
      transparency
      hazard category
      transparency
      hazard name
      Organization transparency
      policy area
      Meta
      assessment type
      organization-level disclosure
      metric
      disclosure_score
      result label
      31 / 100
      sample size
      100
      updated at
      2026-11-09T18:08:05.227Z

      Source specification observations

      1–3 / 3

      openRouter ↗

      field
      context_length
      value
      1310720
      normalized value
      1310720
      scope
      source-model
      source id
      openRouter
      confidence
      1

      openRouter ↗

      field
      knowledge_cutoff
      reported field
      knowledge_cutoff
      value
      2024-08-31
      normalized value
      2024-08-31
      precision
      day
      scope
      source-model
      source id
      openRouter
      confidence
      1

      openRouter ↗

      field
      max_output_tokens
      value
      16384
      normalized value
      16384
      scope
      source-model
      source id
      openRouter
      confidence
      1

      Original sources

      1–17 / 17

      ai.meta.com · Model reference ↗

      Hugging Face · Model reference ↗

      Cloudflare AI · Model page ↗

      DeepInfra · Model page ↗

      Google Cloud · Model page ↗

      Google Cloud · Model page ↗

      OpenRouter · Model page ↗

      GitHub · Source snapshot ↗

      Cloudflare AI · Catalog ↗

      DeepInfra · Catalog ↗

      Epoch AI · Catalog ↗

      GitHub · Catalog ↗

      GitHub · Catalog ↗

      Google Cloud · Catalog ↗

      LMArena · Catalog ↗

      OpenRouter · Catalog ↗

      gorilla.cs.berkeley.edu · Catalog ↗

      Source links

      1–12 / 12

      Explore more models

      Models

      Loading
      EVIDENCE / COMPARISON

      Benchmarks

      Explore every benchmark

      Loading

      These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

      Install TensorTrace