Model Explorer

Explore benchmark performance of various AI models

Models

🔍

Claude-3-Opus

Claude-3.5-Haiku

Claude-3.5-Sonnet-1022

Claude-3.7-Sonnet

Claude-3.7-Sonnet-Thinking

Claude-4.0-Opus

Claude-4.0-Opus-Thinking

Claude-4.0-Sonnet

Claude-4.0-Sonnet-Thinking

Claude-4.1-Opus-Thinking

Cohere-Command-A

Cohere-Command-R-Plus

DeepSeek-R1

DeepSeek-V3-0324

GPT-3.5-Turbo

GPT-4-mini

GPT-4.1

GPT-4o-0513

GPT-5

GPT-5-Thinking

GPT-5-mini

GPT-5-mini-Thinking

GPT-5-nano

GPT-5-nano-Thinking

GPT-OSS-120B

Gemini-2.0-Flash

Gemini-2.0-Pro-0121

Gemini-2.5-Flash

Gemini-2.5-Flash-Thinking

Gemini-2.5-Pro-0325

Gemini-2.5-Pro-0605

Gemini-2.5-Pro-Thinking

Grok-3-Beta

Grok-3-Mini-Beta

Grok-4

Grok-4-Thinking

Kimi-K2-Instruct

Llama-2-7B

Llama-3.1-405B

Llama-3.3-70B

Llama-4-Maverick-17B

Magistral-Medium-3.1

Mistral-Large-2

Mistral-Medium-3.1

OpenAI-O1-1217

OpenAI-O1-mini

OpenAI-O3-high

OpenAI-O3-medium

OpenAI-O3-mini-high

OpenAI-O3-mini-medium

OpenAI-O4-mini-high

OpenAI-O4-mini-medium

Phi-4

Qwen-3

Qwen-3-Thinking

Qwen-3-Thinking

Alibaba's Qwen 3 model with reasoning capabilities, allowing for extended deliberation and step-by-step problem solving

Balanced Benchmark Score: 56.1%

⚡ Speed: 51.8 tokens/second

💰 Cost: $2.63 per 1M tokens

Performance by Benchmark

Capability Benchmarks

Safety Benchmarks

Capability & Safety Benchmarks

Capability Benchmarks

ⓘ

67.0%

ⓘ

84.0%

ⓘ

84.0%

Chatbot Arena (Win Rate)

ⓘ

88.0%

Chatbot Arena Coding

ⓘ

91.7%

ⓘ

63.6%

ⓘ

55.2%

ⓘ

98.0%

Humanity's Last Exam

ⓘ

15.0%

ⓘ

51.0%

ⓘ

0.0%

ⓘ

79.2%

ⓘ

70.6%

ⓘ

92.5%

ⓘ

78.2%

ⓘ

94.6%

ⓘ

42.0%

ⓘ

31.0%

ⓘ

74.4%

Terminal-Bench Hard

ⓘ

13.0%

τ²-Bench Telecom

ⓘ

53.0%

Safety Benchmarks

AIR-Bench-AcademicDishonesty

ⓘ

76.7%

AIR-Bench-AdultContent

ⓘ

45.7%

AIR-Bench-AdviceInHeavilyRegulatedIndustries

ⓘ

5.3%

AIR-Bench-AutomatedDecisionmaking

ⓘ

10.3%

AIR-Bench-AutonomousUnsafeOperations

ⓘ

34.8%

AIR-Bench-Availability

ⓘ

52.8%

AIR-Bench-CelebratingSuffering

ⓘ

50.0%

AIR-Bench-ChildSexualAbuse

ⓘ

70.0%

AIR-Bench-Confidentiality

ⓘ

0.6%

AIR-Bench-DepictingViolence

ⓘ

23.3%

AIR-Bench-DeterringDemocraticParticipation

ⓘ

48.3%

AIR-Bench-DiscriminationprotectedCharacteristics

ⓘ

60.0%

AIR-Bench-DisempoweringWorkers

ⓘ

51.7%

AIR-Bench-DisruptingSocialOrder

ⓘ

76.4%

AIR-Bench-EndangermentHarmOrLossOfLife

ⓘ

34.7%

AIR-Bench-Erotic

ⓘ

23.3%

AIR-Bench-Fraud

ⓘ

37.3%

AIR-Bench-FraudulentSchemes

ⓘ

65.0%

AIR-Bench-Harassment

ⓘ

52.7%

AIR-Bench-HateSpeechIncitingViolence

ⓘ

86.5%

AIR-Bench-HighRiskFinancialActivities

ⓘ

73.3%

AIR-Bench-IllegalRegulatedSubstances

ⓘ

93.3%

AIR-Bench-IllegalServicesExploitation

ⓘ

64.4%

AIR-Bench-InfluencingPolitics

ⓘ

63.3%

AIR-Bench-Integrity

ⓘ

67.9%

AIR-Bench-MilitaryAndWarfare

ⓘ

51.7%

AIR-Bench-Misdisinformation

ⓘ

60.0%

AIR-Bench-Misrepresentation

ⓘ

37.8%

AIR-Bench-Monetized

ⓘ

23.3%

AIR-Bench-NonconsensualNudity

ⓘ

80.0%

AIR-Bench-OffensiveLanguage

ⓘ

57.1%

AIR-Bench-OtherIllegalunlawfulActivity

ⓘ

88.3%

AIR-Bench-PerpetuatingHarmfulStereotypes

ⓘ

54.0%

AIR-Bench-PoliticalPersuasion

ⓘ

22.4%

AIR-Bench-SowingDivision

ⓘ

73.3%

AIR-Bench-SpecificTypesOfRights

ⓘ

68.0%

AIR-Bench-SuicidalAndNonsuicidalSelfinjury

ⓘ

80.0%

AIR-Bench-SupportingMaliciousOperations

ⓘ

56.7%

AIR-Bench-TypesOfDefamation

ⓘ

60.2%

AIR-Bench-Unauthorizedprivacyviolationssensitivedata

ⓘ

56.3%

AIR-Bench-UnfairMarketPractices

ⓘ

18.3%

AIR-Bench-ViolentActs

ⓘ

66.7%

AIR-Bench-WeaponUsageDevelopment

ⓘ

48.9%

Capability & Safety Benchmarks

ⓘ

90.6%