Qwen: Qwen3 30B A3B

Qwen3, the latest generation in the Qwen large language model series, features both dense and mixture-of-experts (MoE) architectures to excel in reasoning, multilingual support, and advanced agent ta…

qwen/qwen3-30b-a3b-04-28

Qwen3, the latest generation in the Qwen large language model series, features both dense and mixture-of-experts (MoE) architectures to excel in reasoning, multilingual support, and advanced agent tasks. Its unique ability to switch seamlessly between a thinking mode for complex reasoning and a non-thinking mode for efficient dialogue ensures versatile, high-quality performance.

Significantly outperforming prior models like QwQ and Qwen2.5, Qwen3 delivers superior mathematics, coding, commonsense reasoning, creative writing, and interactive dialogue capabilities. The Qwen3-30B-A3B variant includes 30.5 billion parameters (3.3 billion activated), 48 layers, 128 experts (8 activated per task), and supports up to 131K token contexts with YaRN, setting a new standard among open-source models.

Specifications & capabilities

Input
text
Output
text
Context window · tokens
40960
Published model ID
Qwen3-30B-A3B
Capabilities
reasoning, structuredOutputs, toolCalling
Description source ↗

Evaluations

1–15 / 15

LLM Stats · GitHub archive

0.804

source id
llmStatsArchive
leaderboard name
AIME 2024 · Raw
dimension
general
metric
AIME 2024 · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
AIME 2024
language
en
modality
text
verified
false
max_score
1
categories
  • math
  • reasoning
created_at
2025-07-19T19:56:11.941652+00:00
paper_link
https://arxiv.org/html/2503.21380v2
updated_at
2025-09-30T00:00:00.000000+00:00
description
American Invitational Mathematics Examination 2024, consisting of 30 challenging mathematical reasoning problems from AIME I and AIME II competitions. Each problem requires an integer answer between 0-999 and tests advanced mathematical reasoning across algebra, geometry, combinatorics, and number theory. Used as a benchmark for evaluating mathematical reasoning capabilities in large language models at Olympiad-level difficulty.
benchmark_id
aime-2024
multilingual
false
analysis_method
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
AIME 2024 · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
3
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.804

source id
llmStatsArchive
leaderboard name
AIME 2024 · Normalized (reported)
dimension
general
metric
AIME 2024 · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
AIME 2024
language
en
modality
text
verified
false
max_score
1
categories
  • math
  • reasoning
created_at
2025-07-19T19:56:11.941652+00:00
paper_link
https://arxiv.org/html/2503.21380v2
updated_at
2025-09-30T00:00:00.000000+00:00
description
American Invitational Mathematics Examination 2024, consisting of 30 challenging mathematical reasoning problems from AIME I and AIME II competitions. Each problem requires an integer answer between 0-999 and tests advanced mathematical reasoning across algebra, geometry, combinatorics, and number theory. Used as a benchmark for evaluating mathematical reasoning capabilities in large language models at Olympiad-level difficulty.
benchmark_id
aime-2024
multilingual
false
analysis_method
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
AIME 2024 · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
3
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.709

source id
llmStatsArchive
leaderboard name
AIME 2025 · Raw
dimension
general
metric
AIME 2025 · Raw
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
AIME 2025
language
en
modality
text
verified
false
max_score
1
categories
  • math
  • reasoning
created_at
2025-09-05T00:00:00.000000+00:00
paper_link
https://arxiv.org/abs/2503.21380
updated_at
2025-09-05T00:00:00.000000+00:00
description
All 30 problems from the 2025 American Invitational Mathematics Examination (AIME I and AIME II), testing olympiad-level mathematical reasoning with integer answers from 000-999. Used as an AI benchmark to evaluate large language models' ability to solve complex mathematical problems requiring multi-step logical deductions and structured symbolic reasoning.
benchmark_id
aime-2025
multilingual
false
analysis_method
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
AIME 2025 · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
2
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.709

source id
llmStatsArchive
leaderboard name
AIME 2025 · Normalized (reported)
dimension
general
metric
AIME 2025 · Normalized (reported)
rank
1
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
AIME 2025
language
en
modality
text
verified
false
max_score
1
categories
  • math
  • reasoning
created_at
2025-09-05T00:00:00.000000+00:00
paper_link
https://arxiv.org/abs/2503.21380
updated_at
2025-09-05T00:00:00.000000+00:00
description
All 30 problems from the 2025 American Invitational Mathematics Examination (AIME I and AIME II), testing olympiad-level mathematical reasoning with integer answers from 000-999. Used as an AI benchmark to evaluate large language models' ability to solve complex mathematical problems requiring multi-step logical deductions and structured symbolic reasoning.
benchmark_id
aime-2025
multilingual
false
analysis_method
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
AIME 2025 · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
2
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.722

source id
llmStatsArchive
leaderboard name
Multi-IF · Normalized (reported)
dimension
general
metric
Multi-IF · Normalized (reported)
rank
2
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
Multi-IF
language
en
modality
text
verified
false
max_score
1
categories
  • reasoning
  • communication
  • language
created_at
2025-07-19T19:56:14.638787+00:00
paper_link
https://arxiv.org/abs/2410.15553
updated_at
2025-07-19T19:56:14.638787+00:00
description
Multi-IF benchmarks LLMs on multi-turn and multilingual instruction following. It expands upon IFEval by incorporating multi-turn sequences and translating English prompts into 7 other languages, resulting in 4,501 multilingual conversations with three turns each. The benchmark reveals that current leading LLMs struggle with maintaining accuracy in multi-turn instructions and shows higher error rates for non-Latin script languages.
benchmark_id
multi-if
multilingual
true
analysis_method
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
Multi-IF · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
4
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.722

source id
llmStatsArchive
leaderboard name
Multi-IF · Raw
dimension
general
metric
Multi-IF · Raw
rank
2
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
Multi-IF
language
en
modality
text
verified
false
max_score
1
categories
  • reasoning
  • communication
  • language
created_at
2025-07-19T19:56:14.638787+00:00
paper_link
https://arxiv.org/abs/2410.15553
updated_at
2025-07-19T19:56:14.638787+00:00
description
Multi-IF benchmarks LLMs on multi-turn and multilingual instruction following. It expands upon IFEval by incorporating multi-turn sequences and translating English prompts into 7 other languages, resulting in 4,501 multilingual conversations with three turns each. The benchmark reveals that current leading LLMs struggle with maintaining accuracy in multi-turn instructions and shows higher error rates for non-Latin script languages.
benchmark_id
multi-if
multilingual
true
analysis_method
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
Multi-IF · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
4
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.658

source id
llmStatsArchive
leaderboard name
GPQA · Normalized (reported)
dimension
general
metric
GPQA · Normalized (reported)
rank
2
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
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/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
11
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.658

source id
llmStatsArchive
leaderboard name
GPQA · Raw
dimension
general
metric
GPQA · Raw
rank
2
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
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/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
11
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.91

source id
llmStatsArchive
leaderboard name
Arena Hard · Normalized (reported)
dimension
general
metric
Arena Hard · Normalized (reported)
rank
3
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
Arena Hard
language
en
modality
text
verified
false
max_score
1
categories
  • general
  • reasoning
  • creativity
created_at
2025-07-19T19:56:14.079874+00:00
paper_link
https://arxiv.org/abs/2406.11939
updated_at
2025-07-19T19:56:14.079874+00:00
description
Arena-Hard-Auto is an automatic evaluation benchmark for instruction-tuned LLMs consisting of 500 challenging real-world prompts curated by BenchBuilder. It includes open-ended software engineering problems, mathematical questions, and creative writing tasks. The benchmark uses LLM-as-a-Judge methodology with GPT-4.1 and Gemini-2.5 as automatic judges to approximate human preference. Arena-Hard achieves 98.6% correlation with human preference rankings and provides 3x higher separation of model performances compared to MT-Bench, making it highly effective for distinguishing between models of similar quality.
benchmark_id
arena-hard
multilingual
false
analysis_method
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
Arena Hard · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
5
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.91

source id
llmStatsArchive
leaderboard name
Arena Hard · Raw
dimension
general
metric
Arena Hard · Raw
rank
3
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
Arena Hard
language
en
modality
text
verified
false
max_score
1
categories
  • general
  • reasoning
  • creativity
created_at
2025-07-19T19:56:14.079874+00:00
paper_link
https://arxiv.org/abs/2406.11939
updated_at
2025-07-19T19:56:14.079874+00:00
description
Arena-Hard-Auto is an automatic evaluation benchmark for instruction-tuned LLMs consisting of 500 challenging real-world prompts curated by BenchBuilder. It includes open-ended software engineering problems, mathematical questions, and creative writing tasks. The benchmark uses LLM-as-a-Judge methodology with GPT-4.1 and Gemini-2.5 as automatic judges to approximate human preference. Arena-Hard achieves 98.6% correlation with human preference rankings and provides 3x higher separation of model performances compared to MT-Bench, making it highly effective for distinguishing between models of similar quality.
benchmark_id
arena-hard
multilingual
false
analysis_method
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
Arena Hard · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
5
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.743

source id
llmStatsArchive
leaderboard name
LiveBench · Normalized (reported)
dimension
general
metric
LiveBench · Normalized (reported)
rank
3
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
LiveBench
language
en
modality
text
verified
false
max_score
1
categories
  • math
  • reasoning
  • general
created_at
2025-09-05T00:00:00.000000+00:00
paper_link
https://arxiv.org/abs/2406.19314
updated_at
2025-09-05T00:00:00.000000+00:00
description
LiveBench is a challenging, contamination-limited LLM benchmark that addresses test set contamination by releasing new questions monthly based on recently-released datasets, arXiv papers, news articles, and IMDb movie synopses. It comprises tasks across math, coding, reasoning, language, instruction following, and data analysis with verifiable, objective ground-truth answers.
benchmark_id
livebench
multilingual
false
analysis_method
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
LiveBench · Normalized (reported)
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
3
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

LLM Stats · GitHub archive

0.743

source id
llmStatsArchive
leaderboard name
LiveBench · Raw
dimension
general
metric
LiveBench · Raw
rank
3
aggregation scope
status
archived
license
CC-BY-4.0
direction
higherIsBetter
conditions
language
en
benchmark
name
LiveBench
language
en
modality
text
verified
false
max_score
1
categories
  • math
  • reasoning
  • general
created_at
2025-09-05T00:00:00.000000+00:00
paper_link
https://arxiv.org/abs/2406.19314
updated_at
2025-09-05T00:00:00.000000+00:00
description
LiveBench is a challenging, contamination-limited LLM benchmark that addresses test set contamination by releasing new questions monthly based on recently-released datasets, arXiv papers, news articles, and IMDb movie synopses. It comprises tasks across math, coding, reasoning, language, instruction following, and data analysis with verifiable, objective ground-truth answers.
benchmark_id
livebench
multilingual
false
analysis_method
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
LiveBench · Raw
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
source_file
data/organizations/qwen/models/qwen3-30b-a3b/benchmarks.json
archive_kind
llmStats
evidence_url
https://qwenlm.github.io/blog/qwen3/
git_revision
872b75f63b8d853c46ed17856424029a0c47b1d2
published_at
2025-10-24T17:47:59Z
snapshot_date
2025-10-24T17:47:59Z
source_revision
6e27aa41f51fa8cb327602349240077d1c7acae9fead8ba2f07db0d549927ddc
publisher_anchor
qwen/qwen3-30b-a3b
metric_population
3
current_catalog_url
https://llm-stats.com/
benchmark_evaluation
true
source model id
qwen3-30b-a3b
Original result ↗

OpenRouter Benchmarks

47%

source id
openRouterBenchmarks
leaderboard name
tau2-bench Verified Airline · Accuracy · DeepInfra
dimension
agentic
metric
tau2-bench Verified Airline · Accuracy
rank
52
sample size
50
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
tau_bench_verified_airline
name
tau2-bench Verified Airline
record_kind
aggregate
provider_name
DeepInfra
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-08T23:09:09.727Z
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
tau2-bench Verified Airline · Accuracy
rank_method
competition rank among reported evaluations in identical conditions, including unlinked observations
sample_size
50
source_file
benchmark-leaderboard.json
source_page
https://openrouter.ai/benchmarks/tau2-bench-airline
measurements
accuracy
0.47
cost_per_task_usd
0.03551402
output_tokens_per_task
20922.04
latency_seconds_per_task
437.81658000000004
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
63
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3-30b-a3b-04-28
source model id
qwen/qwen3-30b-a3b-04-28
Original result ↗

OpenRouter Benchmarks

60.943%

source id
openRouterBenchmarks
leaderboard name
GPQA Diamond · Accuracy · DeepInfra
dimension
knowledge
metric
GPQA Diamond · Accuracy
rank
56
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
DeepInfra
attribution
Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter.
last_run_at
2026-10-08T22:57:53.933Z
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
0.609428
cost_per_task_usd
0.00778269696969697
output_tokens_per_task
15369.68686868687
latency_seconds_per_task
302.9471666666667
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
71
benchmark_evaluation
true
openrouter_permaslug
qwen/qwen3-30b-a3b-04-28
source model id
qwen/qwen3-30b-a3b-04-28
Original result ↗

Published prices

1–8 / 8

qwen-3-30b ↗

$0.10

Inference provider
cheaperInference
Platform
Cheaper Inference
Unit
input_tokens_per_million
Currency
USD
Billing conditions
zdr
false
protocol
chat_completions
service_tier
standard
Billing label
Input tokens (per 1M)

qwen-3-30b ↗

$0.40

Inference provider
cheaperInference
Platform
Cheaper Inference
Unit
output_tokens_per_million
Currency
USD
Billing conditions
zdr
false
protocol
chat_completions
service_tier
standard
Billing label
Output tokens (per 1M)

qwen/qwen3-30b-a3b ↗

$0.12

Inference provider
openRouter
Platform
OpenRouter
Unit
input_tokens_per_million
Currency
USD
Billing conditions

qwen/qwen3-30b-a3b ↗

$0.50

Inference provider
openRouter
Platform
OpenRouter
Unit
output_tokens_per_million
Currency
USD
Billing conditions

alibaba/qwen-3-30b ↗

$0.12

Inference provider
vercelAIGatewayCatalog
Platform
deepinfra
Unit
input_tokens_per_million
Currency
USD
Billing conditions
endpoint
deepinfra | alibaba/qwen-3-30b
model_type
language
quantization
fp8
billing_scope
via Vercel AI Gateway
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

alibaba/qwen-3-30b ↗

$0.50

Inference provider
vercelAIGatewayCatalog
Platform
deepinfra
Unit
output_tokens_per_million
Currency
USD
Billing conditions
endpoint
deepinfra | alibaba/qwen-3-30b
model_type
language
quantization
fp8
billing_scope
via Vercel AI Gateway
Billing label
Output tokens / 1M
Pricing via
Vercel AI Gateway

alibaba/qwen-3-30b ↗

$0.12

Inference provider
vercelAIGatewayCatalog
Platform
Vercel AI Gateway
Unit
input_tokens_per_million
Currency
USD
Billing conditions
model_type
language
quote_scope
gateway_catalog
billing_scope
Vercel public catalog; endpoint prices may differ
Billing label
Input tokens / 1M
Pricing via
Vercel AI Gateway

alibaba/qwen-3-30b ↗

$0.50

Inference provider
vercelAIGatewayCatalog
Platform
Vercel AI Gateway
Unit
output_tokens_per_million
Currency
USD
Billing conditions
model_type
language
quote_scope
gateway_catalog
billing_scope
Vercel public catalog; endpoint prices may differ
Billing label
Output tokens / 1M
Pricing via
Vercel AI Gateway

Provider specifications & access

1–4 / 4

Qwen3-30B-A3B ↗

source model id
alibaba/qwen-3-30b
hosting
vercel-gateway
platforms
  • Vercel AI Gateway
context length
40960
max output tokens
16384
architecture
input_modalities
  • text
output_modalities
  • text
capabilities
  • reasoning
  • structuredOutputs
  • toolCalling

Qwen3-30B-A3B ↗

source model id
publishers/qwen/models/qwen3@qwen3-30b-a3b
hosting
vertex-self-deployed
platforms
  • Google Cloud Vertex AI
architecture
input_modalities
  • text
output_modalities
  • text
capabilities

    qwen-3-30b ↗

    source model id
    qwen-3-30b
    hosting
    gateway
    platforms
    • Cheaper Inference
    context length
    40960
    max output tokens
    16384
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities
    • reasoning
    • structuredOutputs
    • toolCalling

    Qwen3-30B-A3B ↗

    source model id
    Qwen/Qwen3-30B-A3B
    hosting
    siliconflow-hosted
    platforms
    • SiliconFlow
    max output tokens
    131072
    architecture
    input_modalities
    • text
    output_modalities
    • text
    capabilities
    • reasoning
    • structuredOutputs
    • toolCalling

    Safety assessments

    1–1 / 1

    Foundation Model Transparency Index 2025 ↗

    26 / 100

    source id
    stanfordFMTI
    canonical model id
    qwen/qwen3-30b-a3b-04-28
    taxonomy
    transparency
    hazard category
    transparency
    hazard name
    Organization transparency
    policy area
    Alibaba
    assessment type
    organization-level disclosure
    metric
    disclosure_score
    result label
    26 / 100
    sample size
    100
    updated at
    2026-11-09T18:08:05.227Z

    Source specification observations

    1–8 / 8

    openRouter ↗

    field
    context_length
    value
    40960
    normalized value
    40960
    scope
    source-model
    source id
    openRouter
    confidence
    1

    siliconFlow ↗

    field
    context_length
    value
    131K
    scope
    source-model
    source id
    siliconFlow
    confidence
    1

    vercelAIGatewayCatalog ↗

    field
    context_length
    value
    40960
    normalized value
    40960
    scope
    source-model
    source id
    vercelAIGatewayCatalog
    confidence
    1

    openRouter ↗

    field
    knowledge_cutoff
    reported field
    knowledge_cutoff
    value
    2025-03-31
    normalized value
    2025-03-31
    precision
    day
    scope
    source-model
    source id
    openRouter
    confidence
    1

    vercelAIGatewayCatalog ↗

    field
    knowledge_cutoff
    reported field
    knowledge_cutoff
    value
    2025-04
    normalized value
    2025-04
    precision
    month
    scope
    source-model
    source id
    vercelAIGatewayCatalog
    confidence
    1

    openRouter ↗

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

    siliconFlow ↗

    field
    max_output_tokens
    value
    131K
    scope
    source-model
    source id
    siliconFlow
    confidence
    1

    vercelAIGatewayCatalog ↗

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

    Original sources

    1–20 / 21

    Qwen · Model reference ↗

    SiliconFlow · Model page ↗

    Hugging Face · Model repository ↗

    Google Cloud · Model page ↗

    OpenRouter · Model page ↗

    SiliconFlow · Playground ↗

    Vercel AI Gateway · Model page ↗

    Google Cloud · Model documentation ↗

    SiliconFlow · API documentation ↗

    arXiv · Paper · 2309.00071 ↗

    arXiv · Paper · 2505.09388 ↗

    api.cheaperinference.com · Source snapshot ↗

    Epoch AI · Catalog ↗

    GitHub · Catalog ↗

    GitHub · Catalog ↗

    Google Cloud · Catalog ↗

    Hugging Face · Catalog ↗

    OpenRouter · Catalog ↗

    OpenRouter · Catalog ↗

    SiliconFlow · Catalog ↗

    Source links

    1–11 / 11

    Hugging Face model card

    1–1 / 1

    Qwen/Qwen3-30B-A3B ↗

    downloads
    1135065
    likes
    984
    pipeline tag
    text-generation
    library name
    transformers
    License
    apache-2.0

    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