Model Leaderboard

Compare AI models by capability and cost-effectiveness

Popular Comparisons

Custom Comparison Browse All Models

Logical Reasoning

71/251 models

HLE: Complex reasoning and problem-solving

Use Cases: Complex decision-making, multi-step analysis, logical reasoning

1Claude Fable 5.159.1

2Claude Fable 555.5

3Claude Opus 554.9

4GPT-6 Astra54.7

5GPT-5.649.5

6Claude Opus 4.848.7

7Gemini 3.7 Flash47.9

8Gemini 3.8 Flash47.8

9Gemini 3.1 Pro Preview47.0

10GPT-5.545.8

11GPT-5.443.7

12GPT-5.5 Pro43.1

13Qwen3.8 Max43.0

14GPT-5.6 Terra42.9

15Grok 4.642.9

Knowledge Q&A

62/251 models

MMLU Pro: Broad knowledge assessment

Use Cases: Expert Q&A, fact-checking, educational tutoring

1Claude Fable 5.171.8

2GPT-6 Astra71.7

3Claude Fable 571.7

4Claude Opus 568.5

5Gemini 3.1 Pro Preview66.0

6Grok 4.665.3

7Gemini 3.8 Flash64.8

8Claude Opus 4.864.4

9Claude Opus 4.763.6

10Gemini 3.7 Flash63.3

11Grok 4.562.6

12Gemini 3.6 Flash61.0

13GPT-5.661.0

14Gemini 3.5 Flash60.6

15GPT-5.560.3

Scientific Research

74/251 models

GPQA: Graduate-level science questions

Use Cases: Academic research, scientific writing, experiment design

1GPT-6 Astra96.1

2Gemini 3.8 Flash95.3

3Grok 4.694.9

4Gemini 3.7 Flash94.5

5GPT-5.694.1

6Gemini 3.1 Pro Preview94.1

7Claude Fable 5.193.7

8GPT-5.593.5

9Claude Opus 593.2

10Grok 4.593.1

11DeepSeek V4 Pro92.8

12Gemini 3.6 Flash92.8

13Qwen3.8 Max92.7

14Claude Fable 592.6

15GPT-5.6 Terra92.5

AI Agent

41/251 models

Tau2: Autonomous task completion

Use Cases: Automated workflows, multi-tool invocation, complex task decomposition

1Qwen3.8 Max51.3

2Grok 4.650.7

3Claude Fable 5.147.2

4GLM-5.3-Flash47.2

5Gemini 3.8 Flash44.9

6GPT-5.644.3

7Grok 4.542.1

8Claude Opus 542.1

9GPT-6 Astra41.4

10GPT-5.6 Terra40.2

11GPT-5.439.6

12DeepSeek V4 Pro39.6

13DeepSeek V4 Flash39.4

14GPT-5.539.0

15Claude Fable 538.1

SciCode

55/251 models

SciCode: Scientific coding challenges

Use Cases: Scientific computing, research code, data analysis scripts

1Claude Fable 5.163.1

2Claude Fable 561.0

3Gemini 3.1 Pro Preview58.7

4Gemini 3.7 Flash57.2

5GPT-5.657.1

6Gemini 3.8 Flash56.6

7GPT-6 Astra56.5

8Grok 4.656.5

9Claude Opus 556.4

10GPT-5.6 Terra55.0

11Grok 4.555.0

12Claude Opus 4.854.4

13Claude Sonnet 554.3

14Gemini 3.5 Flash53.9

15GPT-5.6 Luna53.6

Terminal

66/251 models

Terminal-Bench 2.1: agentic coding & terminal operations

Use Cases: Agentic coding, shell scripting, DevOps automation

1Claude Fable 5.191.4

2Claude Opus 589.1

3GPT-6 Astra88.4

4Grok 4.688.4

5GPT-5.688.0

6GPT-5.6 Terra88.0

7Gemini 3.8 Flash87.6

8Gemini 3.7 Flash85.8

9Claude Fable 584.6

10Claude Opus 4.884.6

11GPT-5.584.3

12GLM-5.3-Flash84.3

13Claude Opus 4.783.1

14Grok 4.581.6

15Qwen3.8 Max81.3

Instruction

53/251 models

IFEval: Instruction following accuracy

Use Cases: Precise task execution, format compliance, constraint adherence

1o3-mini93.4

2Grok 4.381.3

3Gemini 3 Flash Preview78.0

4GPT-5.2-Codex77.6

5Gemini 3.1 Flash-Lite Preview77.2

6Gemini 3.1 Pro Preview77.1

7GLM-5.176.3

8Gemini 3.5 Flash76.3

9GPT-5.575.9

10GPT-5.4 Nano75.9

11GPT-5.275.4

12GPT-5 Mini75.4

13GPT-5 Codex74.1

14GPT-5.473.9

15GPT-5.4 Mini73.3