Meta: Llama 3.2 3B Instruct

Llama 3.2 3B is a 3-billion-parameter multilingual large language model, optimized for advanced natural language processing tasks like dialogue generation, reasoning, and summarization. Designed with…

meta-llama/llama-3.2-3b-instruct

Llama 3.2 3B is a 3-billion-parameter multilingual large language model, optimized for advanced natural language processing tasks like dialogue generation, reasoning, and summarization. Designed with the latest transformer architecture, it supports eight languages, including English, Spanish, and Hindi, and is adaptable for additional languages.

Trained on 9 trillion tokens, the Llama 3.2 3B model excels in instruction-following, complex reasoning, and tool use. Its balanced performance makes it ideal for applications needing accuracy and efficiency in text generation across multilingual settings.

Click here for the [original model card](https://github.com/meta-llama/llama-models/blob/main/models/llama3_2/MODEL_CARD.md).

Usage of this model is subject to [Meta's Acceptable Use Policy](https://www.llama.com/llama3/use-policy/).

Specifications & capabilities

Input
text
Output
text
Context window · tokens
131072
Knowledge cutoff
2023-12-31
Published model ID
Llama-3.2-3B-Instruct
Capabilities
structuredOutputs
Description source ↗

Evaluations

1–20 / 69

LLM Stats · GitHub archive

0.401

source id
llmStatsArchive
leaderboard name
Open-rewrite · Raw
dimension
general
metric
Open-rewrite · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
Open-rewrite
language
en
modality
text
verified
false
max_score
1
categories
  • language
  • writing
created_at
2025-07-19T19:56:14.435616+00:00
paper_link
https://arxiv.org/abs/2305.15685
updated_at
2025-07-19T19:56:14.435616+00:00
description
OpenRewriteEval is a benchmark for evaluating open-ended rewriting of long-form texts, covering a wide variety of rewriting types expressed through natural language instructions including formality, expansion, conciseness, paraphrasing, and tone and style transfer.
benchmark_id
open-rewrite
multilingual
false
analysis_method
0-shot, micro_avg/rougeL
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
Open-rewrite · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-3.2-3b-instruct/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-3.2-3B-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-3.2-3b-instruct
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-3.2-3b-instruct
Original result ↗

LLM Stats · GitHub archive

0.401

source id
llmStatsArchive
leaderboard name
Open-rewrite · Normalized (reported)
dimension
general
metric
Open-rewrite · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
Open-rewrite
language
en
modality
text
verified
false
max_score
1
categories
  • language
  • writing
created_at
2025-07-19T19:56:14.435616+00:00
paper_link
https://arxiv.org/abs/2305.15685
updated_at
2025-07-19T19:56:14.435616+00:00
description
OpenRewriteEval is a benchmark for evaluating open-ended rewriting of long-form texts, covering a wide variety of rewriting types expressed through natural language instructions including formality, expansion, conciseness, paraphrasing, and tone and style transfer.
benchmark_id
open-rewrite
multilingual
false
analysis_method
0-shot, micro_avg/rougeL
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
Open-rewrite · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-3.2-3b-instruct/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-3.2-3B-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-3.2-3b-instruct
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-3.2-3b-instruct
Original result ↗

LLM Stats · GitHub archive

0.198

source id
llmStatsArchive
leaderboard name
InfiniteBench/En.QA · Normalized (reported)
dimension
general
metric
InfiniteBench/En.QA · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
InfiniteBench/En.QA
language
en
modality
text
verified
false
max_score
1
categories
  • long_context
created_at
2025-07-19T19:56:14.457927+00:00
paper_link
https://arxiv.org/abs/2402.13718
updated_at
2025-07-19T19:56:14.457927+00:00
description
InfiniteBench English Question Answering variant - first LLM benchmark featuring average data length surpassing 100K tokens for evaluating long-context capabilities with 12 tasks spanning diverse domains
benchmark_id
infinitebench-en.qa
multilingual
false
analysis_method
0-shot, longbook_qa/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
InfiniteBench/En.QA · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-3.2-3b-instruct/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-3.2-3B-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-3.2-3b-instruct
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-3.2-3b-instruct
Original result ↗

LLM Stats · GitHub archive

0.198

source id
llmStatsArchive
leaderboard name
InfiniteBench/En.QA · Raw
dimension
general
metric
InfiniteBench/En.QA · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
InfiniteBench/En.QA
language
en
modality
text
verified
false
max_score
1
categories
  • long_context
created_at
2025-07-19T19:56:14.457927+00:00
paper_link
https://arxiv.org/abs/2402.13718
updated_at
2025-07-19T19:56:14.457927+00:00
description
InfiniteBench English Question Answering variant - first LLM benchmark featuring average data length surpassing 100K tokens for evaluating long-context capabilities with 12 tasks spanning diverse domains
benchmark_id
infinitebench-en.qa
multilingual
false
analysis_method
0-shot, longbook_qa/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
InfiniteBench/En.QA · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-3.2-3b-instruct/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-3.2-3B-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-3.2-3b-instruct
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-3.2-3b-instruct
Original result ↗

LLM Stats · GitHub archive

0.19

source id
llmStatsArchive
leaderboard name
TLDR9+ (test) · Raw
dimension
general
metric
TLDR9+ (test) · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
TLDR9+ (test)
language
en
modality
text
verified
false
max_score
1
categories
  • summarization
  • language
created_at
2025-07-19T19:56:14.439927+00:00
paper_link
https://arxiv.org/abs/2110.01159
updated_at
2025-07-19T19:56:14.439927+00:00
description
A large-scale summarization dataset containing over 9 million training instances extracted from Reddit, designed for extreme summarization (generating one-sentence summaries with high compression and abstraction). More than twice larger than previously proposed datasets.
benchmark_id
tldr9+-(test)
multilingual
false
analysis_method
1-shot, rougeL
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
TLDR9+ (test) · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-3.2-3b-instruct/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-3.2-3B-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-3.2-3b-instruct
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-3.2-3b-instruct
Original result ↗

LLM Stats · GitHub archive

0.19

source id
llmStatsArchive
leaderboard name
TLDR9+ (test) · Normalized (reported)
dimension
general
metric
TLDR9+ (test) · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
TLDR9+ (test)
language
en
modality
text
verified
false
max_score
1
categories
  • summarization
  • language
created_at
2025-07-19T19:56:14.439927+00:00
paper_link
https://arxiv.org/abs/2110.01159
updated_at
2025-07-19T19:56:14.439927+00:00
description
A large-scale summarization dataset containing over 9 million training instances extracted from Reddit, designed for extreme summarization (generating one-sentence summaries with high compression and abstraction). More than twice larger than previously proposed datasets.
benchmark_id
tldr9+-(test)
multilingual
false
analysis_method
1-shot, rougeL
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
TLDR9+ (test) · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/meta/models/llama-3.2-3b-instruct/benchmarks.json
archive_kind
llmStats
evidence_url
https://huggingface.co/meta-llama/Llama-3.2-3B-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-3.2-3b-instruct
metric_population
1
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
llama-3.2-3b-instruct
Original result ↗

OpenRouter Benchmarks

28.62%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · Cloudflare
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
17
sample size
198
aggregation scope
status
published
harness
url
https://github.com/OpenRouterTeam/benchmark-harness
name
OpenRouter benchmark harness
direction
higherIsBetter
run_count
1
conditions
window
newest 90 days; successful visible nonzero runs
dataset
benchmark-leaderboard
benchmark
id
gpqa_diamond
name
GPQA Diamond
record_kind
aggregate
provider_name
Cloudflare
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-13T18:56:05.001Z
license_url
https://creativecommons.org/licenses/by/4.0/
methodology
Rows aggregate the benchmark-runs dataset exactly as the leaderboard pages do, over each benchmark's non-zero-score runs from the newest 90 days: accuracy is the task-weighted mean of per-run accuracy, accuracy_std_dev is the population standard deviation of per-run accuracy (null with a single run), and the per-task averages divide summed cost, generation time, and output tokens by summed tasks. Rows are ordered by accuracy descending, then model name.
metric_kind
GPQA Diamond · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
198
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/gpqa-diamond
measurements
accuracy
0.286195
cost_per_task_usd
0.0005941969696969697
output_tokens_per_task
1642.1363636363637
latency_seconds_per_task
14.998156565656565
published_at
2026-11-16T01:46:02.653Z
snapshot_date
2026-11-16
source_revision
7ccaf17664252a3053dcb7df8ad0a46e72954a3b4524ddc1c8a7ef8a41b15568
transformations
TensorTrace preserves source results, converts latency to seconds and derives ranks within matching conditions. Individual runs have no derived rank.
metric_population
18
benchmark_evaluation
true
openrouter_permaslug
meta-llama/llama-3.2-3b-instruct
source model id
meta-llama/llama-3.2-3b-instruct
Original result ↗

LMArena

1,109.865

source id
lmarena
leaderboard name
LMArena Text Style Control German
dimension
style
metric
Arena rating
rank
302
sample size
232
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1068.7704047691923
confidence upper
1150.9596990759055
Original result ↗

LMArena

1,055.989

source id
lmarena
leaderboard name
LMArena Text German
dimension
general
metric
Arena rating
rank
302
sample size
232
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1015.5243730292343
confidence upper
1096.4534646612512
Original result ↗

LMArena

1,089.636

source id
lmarena
leaderboard name
LMArena Text Expert
dimension
general
metric
Arena rating
rank
339
sample size
499
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1065.4484337107501
confidence upper
1113.822680764903
Original result ↗

LMArena

1,104.319

source id
lmarena
leaderboard name
LMArena Text Industry Medicine And Healthcare
dimension
general
metric
Arena rating
rank
342
sample size
441
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1074.3256505705483
confidence upper
1134.3120678198986
Original result ↗

LMArena

1,178.863

source id
lmarena
leaderboard name
LMArena Text Style Control Expert
dimension
style
metric
Arena rating
rank
343
sample size
499
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1153.924560086592
confidence upper
1203.8006030252504
Original result ↗

LMArena

1,202.891

source id
lmarena
leaderboard name
LMArena Text Style Control Industry Medicine And Healthcare
dimension
style
metric
Arena rating
rank
346
sample size
441
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1172.235316195937
confidence upper
1233.546505471072
Original result ↗

LMArena

1,128.549

source id
lmarena
leaderboard name
LMArena Text Industry Mathematical
dimension
general
metric
Arena rating
rank
346
sample size
980
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1111.8178617484543
confidence upper
1145.2792443829694
Original result ↗

LMArena

1,179.621

source id
lmarena
leaderboard name
LMArena Text English
dimension
general
metric
Arena rating
rank
347
sample size
4157
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1170.0153060184534
confidence upper
1189.2258448090843
Original result ↗

LMArena

1,125.881

source id
lmarena
leaderboard name
LMArena Text Math
dimension
general
metric
Arena rating
rank
347
sample size
1136
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1110.196596931938
confidence upper
1141.5652055553498
Original result ↗

LMArena

1,165.325

source id
lmarena
leaderboard name
LMArena Text Style Control Math
dimension
style
metric
Arena rating
rank
349
sample size
1136
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1149.3906252843444
confidence upper
1181.2593801162038
Original result ↗

LMArena

1,140.743

source id
lmarena
leaderboard name
LMArena Text Hard Prompts English
dimension
general
metric
Arena rating
rank
349
sample size
1377
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1125.7886633042028
confidence upper
1155.6963811310043
Original result ↗

LMArena

1,115.291

source id
lmarena
leaderboard name
LMArena Text Industry Legal And Government
dimension
general
metric
Arena rating
rank
349
sample size
506
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1087.3488608918533
confidence upper
1143.232979120907
Original result ↗

LMArena

1,174.538

source id
lmarena
leaderboard name
LMArena Text Style Control Industry Mathematical
dimension
style
metric
Arena rating
rank
350
sample size
980
evaluated at
2026-11-09T00:00:00.000Z
source model id
Llama 3.2 3B Instruct
confidence lower
1157.3999103456456
confidence upper
1191.6751882426413
Original result ↗

Published prices

1–6 / 6

@cf/meta/llama-3.2-3b-instruct ↗

$0.05

Inference provider
cloudflareWorkersAI
Platform
Cloudflare Workers AI
Unit
input_tokens_per_million
Currency
USD
Billing conditions
Billing label
per M input tokens

@cf/meta/llama-3.2-3b-instruct ↗

$0.34

Inference provider
cloudflareWorkersAI
Platform
Cloudflare Workers AI
Unit
output_tokens_per_million
Currency
USD
Billing conditions
Billing label
per M output tokens

Provider specifications & access

1–3 / 3

llama-3.2-3b-instruct ↗

source model id
@cf/meta/llama-3.2-3b-instruct
hosting
cloudflare-hosted
platforms
  • Cloudflare Workers AI
context length
80000
architecture
input_modalities
  • text
output_modalities
  • text
capabilities

    Llama-3.2-3B-Instruct ↗

    source model id
    publishers/hf-meta-llama/models/llama-3.2-3b-instruct@001
    hosting
    vertex-self-deployed
    platforms
    • Google Cloud Vertex AI
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities

      Llama-3.2-3B-Instruct ↗

      source model id
      publishers/meta/models/llama3-2@llama-3.2-3b-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-3.2-3b-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
        131072
        normalized value
        131072
        scope
        source-model
        source id
        openRouter
        confidence
        1

        openRouter ↗

        field
        knowledge_cutoff
        reported field
        knowledge_cutoff
        value
        2023-12-31
        normalized value
        2023-12-31
        precision
        day
        scope
        source-model
        source id
        openRouter
        confidence
        1

        openRouter ↗

        field
        max_output_tokens
        value
        72000
        normalized value
        72000
        scope
        source-model
        source id
        openRouter
        confidence
        1

        Original sources

        1–19 / 19

        Hugging Face · Model repository ↗

        Cloudflare AI · Model page ↗

        DeepInfra · Model page ↗

        Google Cloud · Model page ↗

        Google Cloud · Model page ↗

        OpenRouter · Model page ↗

        arXiv · Paper · 2204.05149 ↗

        arXiv · Paper · 2405.16406 ↗

        GitHub · Source snapshot ↗

        Cloudflare AI · Catalog ↗

        DeepInfra · Catalog ↗

        GitHub · Catalog ↗

        GitHub · Catalog ↗

        Google Cloud · Catalog ↗

        Hugging Face · Catalog ↗

        Hugging Face · Catalog ↗

        LMArena · Catalog ↗

        OpenRouter · Catalog ↗

        OpenRouter · Catalog ↗

        Source links

        1–11 / 11

        Hugging Face model card

        1–1 / 1

        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