Qwen3 8B (Reasoning)

April 28, 2025Last synced Apr 7, 2026, 4:00 PM

Blended Price

$0.66/M

Input Price

$0.18/M

Output Price

$2.10/M

Speed

88 tok/s

TTFT

0.94s

Benchmark Scores

Intelligence Index

External Benchmarks

AIME 2024benchmark_matrix

How Qwen3 8B (Reasoning) Compares

Axes

X Axis

Y Axis

Bubble Size

Filters

Blended Price (USD)$0.00 – $30.0

AgMoBench Overall≥ 3.5

Providers

ai21-labsalibabaanthropicawsazurebaiducoheredeepseek

Show quadrants

anthropic

openai

google

meta

mistral

nvidia

kimi

xai

azure

deepseek

aws

cohere

baidu

ai21-labs

zai

reka-ai

xiaomi

alibaba

minimax

ibm

Bubble size = Context Window

16384.00

2000000.00

Compare with other models

GPQA Diamond

0.6 / 100

LiveCodeBench

0.4 / 100

Terminal-Bench Hard

0.0 / 100

τ²-Bench

0.3 / 100

AIME 2025 (Matrix)benchmark_matrix

67.3

Arena-Hard Autobenchmark_matrix

76.0

Codeforces Ratingbenchmark_matrix

1850.0

GPQA Diamond (Matrix)benchmark_matrix

63.3

GSM8Kbenchmark_matrix

89.8

HumanEvalbenchmark_matrix

82.0

IFBench (Matrix)benchmark_matrix

35.0

IFEvalbenchmark_matrix

80.0

LiveCodeBench (Matrix)benchmark_matrix

48.0

MATH-500 (Matrix)benchmark_matrix

94.0

MMLUbenchmark_matrix

76.9

MMLU-Pro (Matrix)benchmark_matrix

56.7

AA-Omniscience AccuracyPredicted

39.1 / 100

AA-Omniscience Hallucination RatePredicted

94.0 / 100

Aider PolyglotPredicted

22.9 / 100

AlpacaEval 2.0 LCPredicted

49.9 / 100

AlpacaEval 2.0 RawPredicted

39.0 / 100

ARC-AGI-1Predicted

2.0 / 100

ARC-AGI-1 Cost per TaskPredicted

0.0

ARC-AGI-2Predicted

0.0 / 100

ARC-AGI-2 Cost per TaskPredicted

0.0

BigCodeBench CompletePredicted

59.7 / 100

BigCodeBench InstructPredicted

49.6 / 100

AA Intelligence Index (Matrix)Predicted

68.4

AA Long Context Reasoning (Matrix)Predicted

69.1

BrowseCompPredicted

52.9

BRUMO 2025Predicted

74.8

CMIMC 2025Predicted

61.3

HLE (Matrix)Predicted

18.5

HMMT Feb 2025Predicted

33.3

HMMT Nov 2025Predicted

87.4

IMO 2025Predicted

8.5

MathArena Apex 2025Predicted

0.4

MMMU-ProPredicted

80.9

MRCR v2Predicted

73.0

OSWorldPredicted

35.2

SimpleQAPredicted

17.4

SMT 2025Predicted

72.6

SWE-bench ProPredicted

33.9

Tau-Bench Telecom (Matrix)Predicted

93.8

Terminal-Bench 2.0Predicted

12.8

Terminal-Bench 1.0Predicted

11.8

USAMO 2025Predicted

4.8

Video-MMUPredicted

80.2

browsecompPredicted

48.9

BullshitBenchPredicted

9.7 / 100

Aider PolyglotPredicted

0.0

Apex AgentsPredicted

1.9

Arc Agi 2Predicted

0.0

BIG-Bench HardPredicted

3.0

CAD-EvalPredicted

6.3

Chess PuzzlesPredicted

0.1

CyBenchPredicted

0.1

DeepResearchBenchPredicted

0.3

FictionLiveBenchPredicted

0.3

GeoBenchPredicted

0.0

GSM8K (Epoch)Predicted

17.8

HellaSwagPredicted

35.5

Lech Mazur WritingPredicted

7.0

METR Time HorizonsPredicted

0.2

OTIS Mock AIME 2024–2025Predicted

0.1

PosttrainbenchPredicted

0.0

SimpleQA Verified (Epoch)Predicted

0.2

The Agent CompanyPredicted

1.4

TriviaQAPredicted

2.3

WinoGrandePredicted

0.9

FrontierMathPredicted

9.0 / 100

GAIA Level 1Predicted

40.2

GAIA Level 2Predicted

43.6

GAIA Level 3Predicted

0.7

GAIAPredicted

26.5 / 100

LegalBenchPredicted

87.9 / 100

LiveBench CodingPredicted

61.7 / 100

LiveBench Data AnalysisPredicted

34.7 / 100

LiveBench LanguagePredicted

42.7 / 100

LiveBench MathPredicted

50.7 / 100

LiveBench OverallPredicted

36.5 / 100

LiveBench ReasoningPredicted

27.9 / 100

LongBench v2 EasyPredicted

44.8

LongBench v2 HardPredicted

36.3

LongBench v2Predicted

36.0 / 100

MathVistaPredicted

55.7 / 100

MLE-benchPredicted

12.4 / 100

MMMUPredicted

67.3 / 100

MMTU Table UnderstandingPredicted

49.1 / 100

MT-BenchPredicted

7.9 / 10

NoLiMa (NIAH)Predicted

94.3 / 100

OCRBench v2Predicted

79.7 / 100

Open LLM AveragePredicted

44.0 / 100

Open LLM: BBHPredicted

69.0 / 100

Open LLM: GPQAPredicted

35.8 / 100

Open LLM: IFEvalPredicted

74.6 / 100

Open LLM: MATH Level 5Predicted

47.5 / 100

Open LLM: MMLU-PROPredicted

55.2 / 100

Open LLM: MUSRPredicted

45.5 / 100

RE-BenchPredicted

0.1

SimpleBenchPredicted

24.8 / 100

simpleqaPredicted

17.7

SWE-bench LitePredicted

15.3 / 100

SWE-bench VerifiedPredicted

32.8 / 100

tau-bench RetailPredicted

71.5 / 100

WebArenaPredicted

18.5 / 100

WeirdMLPredicted

16.2 / 100

WildBenchPredicted

47.1

BFCL (Berkeley Function Calling)bfcl

42.6

EQ-Bench 3eqbench

1251.0

Vectara Factual Consistencyvectara_hallucination

95.2 / 100

Vectara Hallucination Ratevectara_hallucination

4.8 / 100