HumanEval Leaderboard

OpenAI HumanEval benchmark measuring Python code generation from function docstrings.

Data from LayerLens

As of November 21, 2026, the top-scoring model on HumanEval is Claude Sonnet 4.5 at 97.6%, followed by Claude Sonnet 4.5 at 97.0% and Claude Opus 4.5 at 97.0%. 43 models have been evaluated on this benchmark.

Last updated: November 21, 2026

8 Ways to Use Fewer Tokens

Models

43

Best Score

97.6

Average

89.2

Std Dev

10.4

Categories
Computer Science and Programming
Sponsored The Grid | Spot Priced LLM API
The Grid | Spot Priced LLM API — Why pay list price when providers will bid for your API usage? Up to 80% off.
View spot prices
Provider
Model
Input $/M
Output $/M
HumanEval
Actions
$3.000
$15.000
97.6
$3.000
$15.000
97.0
$5.000
$25.000
97.0
$5.000
$25.000
97.0
$5.000
$25.000
97.0
$5.000
$25.000
97.0
$0.600
$1.920
97.0
$0.600
$1.920
97.0
$1.000
$5.000
96.3
$0.390
$0.900
96.3
$0.064
$0.224
95.7
$0.064
$0.224
95.7
$0.625
$5.000
94.5
$1.000
$5.000
93.9
$0.209
$0.310
93.9
$2.000
$8.000
93.3
$0.120
$0.500
93.3
$0.120
$0.500
93.3
$0.500
$2.150
93.3
$0.060
$0.240
93.3
$0.060
$0.240
93.3
$0.260
$1.560
93.3
$0.300
$0.900
92.7
$0.000
$0.000
92.1
$0.060
$0.400
92.1
$0.060
$0.400
92.1
$0.900
$0.900
90.2
$0.900
$0.900
90.2
$0.500
$1.500
90.2
$0.440
$2.200
88.4
$0.014
$0.028
87.2
$0.080
$0.160
85.4
$0.440
$2.200
85.4
$0.070
$0.140
84.8
$0.200
$0.600
84.8
$0.075
$0.200
83.5
$2.500
$10.000
82.9
$0.288
$0.320
82.3
$0.100
$0.300
81.1
$0.060
$0.240
78.0
$0.800
$3.200
76.8
$0.000
$0.000
51.2
$0.030
$0.050
47.6

Pricing from OpenRouter. Benchmarks from Artificial Analysis.

About HumanEval

OpenAI HumanEval benchmark measuring Python code generation from function docstrings.

This leaderboard shows all models with HumanEval benchmark scores, ranked from highest to lowest. Pricing data is included to help you compare performance against cost.

Frequently Asked Questions

OpenAI HumanEval benchmark measuring Python code generation from function docstrings.
As of November 21, 2026, Claude Sonnet 4.5 leads the HumanEval leaderboard with a score of 97.6. Rankings change as new models are released and evaluated.
Currently 43 models have been evaluated on HumanEval, with an average score of 89.2 and standard deviation of 10.4.
Benchmark scores are updated when new evaluations are published by our data sources (Artificial Analysis and LayerLens). Pricing data is refreshed daily from OpenRouter.