Current Top 3
#1GPT-5.4
OpenAI
$2.50 / 1M
$15.00 / 1M
#2Claude Opus 4.6
Anthropic
$5.00 / 1M
$25.00 / 1M
#3Claude Sonnet 4.6
Anthropic
$3.00 / 1M
$15.00 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
Open← Back to home
Scenario Leaderboards
Start with the overall LLM leaderboard, then drill into scenario-specific rankings. Each list blends community votes, public-solution adoption, editorial picks, and pricing signals.
🦀
AI Coding Agent Benchmark
PinchBench · 24 models · 23 tasks · Success rate + API pricing comparison
View benchmark →
Current Top 3
#1Gemini 2.0 Flash Lite
Google AI
$0.07 / 1M
$0.30 / 1M
#2Gemini 2.0 Flash
Google AI
$0.10 / 1M
$0.40 / 1M
#3DeepSeek Reasoner
DeepSeek
$0.28 / 1M
$0.42 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
OpenCurrent Top 3
#1Claude Opus 4.6
Anthropic
$5.00 / 1M
$25.00 / 1M
#2Claude Sonnet 4.6
Anthropic
$3.00 / 1M
$15.00 / 1M
#3Gemini 2.0 Flash Lite
Google AI
$0.07 / 1M
$0.30 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
OpenCurrent Top 3
#1Claude 3.5 Sonnet
Anthropic
$3.00 / 1M
$15.00 / 1M
#2Gemini 2.0 Flash Lite
Google AI
$0.07 / 1M
$0.30 / 1M
#3Gemini 2.0 Flash
Google AI
$0.10 / 1M
$0.40 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
OpenCurrent Top 3
#1Claude 3.5 Sonnet
Anthropic
$3.00 / 1M
$15.00 / 1M
#2DeepSeek V3
DeepSeek
$0.27 / 1M
$1.10 / 1M
#3Gemini 2.0 Flash Lite
Google AI
$0.07 / 1M
$0.30 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
OpenCurrent Top 3
#1DeepSeek V3
DeepSeek
$0.27 / 1M
$1.10 / 1M
#2Gemini 2.0 Flash Lite
Google AI
$0.07 / 1M
$0.30 / 1M
#3Gemini 2.0 Flash
Google AI
$0.10 / 1M
$0.40 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
OpenCurrent Top 3
#1Claude 3.5 Sonnet
Anthropic
$3.00 / 1M
$15.00 / 1M
#2Gemini 2.0 Flash Lite
Google AI
$0.07 / 1M
$0.30 / 1M
#3Gemini 2.0 Flash
Google AI
$0.10 / 1M
$0.40 / 1M
Open the full leaderboard, vote, and jump into solutions using each model.
Open