Deterministic Stack Engine

Best AI Stack for frontier agents

Coding + preference Elo for computer-use and multi-step tools.

Objective Weight Distribution

SWE-bench:35%Elo:25%Speed:15%TTFT:10%Out $:15%

Recommended PickScore 79.2 / 100

Gemini 3.7 Flash

Google · Closed flagship

SWE-bench: 72.4% (Top 21%)Elo: 1,530 (Top 23%)Speed: 215 tok/s (Top 1%)TTFT: 82 ms (Top 1%)

Output: $1.875/1MSWE-bench: 72.4%

View full model fact sheet

Full Stack Leaderboard

Ranked alternatives optimized for frontier agents based on objective benchmark weighting.

#2Gemini 3 Flash(Google)

Score 76.9

SWE-bench75.8%Elo1,495Speed190 tok/sTTFT95 ms

Rank #2 in stackvs #1

#3DeepSeek V4 Flash(DeepSeek)

Score 73.5

SWE-bench71.2%Elo1,490Speed142 tok/sTTFT165 ms

Rank #3 in stackvs #1

#4Gemini 3.6 Flash(Google)

Score 72.0

SWE-bench70.8%Elo1,506Speed198 tok/sTTFT92 ms

Rank #4 in stackvs #1

#5GLM-5.3(Zhipu)

Score 71.2

SWE-bench76.4%Elo1,558Speed90 tok/sTTFT255 ms

Rank #5 in stackvs #1

#6Gemini 3.6 Pro(Google)

Score 71.0

SWE-bench74.6%Elo1,570Speed102 tok/sTTFT205 ms

Rank #6 in stackvs #1

#7OpenAI: o3 Mini(OpenAI)

Score 70.5

SWE-bench78.5%Elo1,560Speed92 tok/sTTFT290 ms

Rank #7 in stackvs #1

#8Claude Sonnet 5(Anthropic)

Score 70.3

SWE-bench73.8%Elo1,556Speed118 tok/sTTFT175 ms

Rank #8 in stackvs #1

#9GPT-5.6 Terra(OpenAI)

Score 67.3

SWE-bench70.4%Elo1,548Speed128 tok/sTTFT160 ms

Rank #9 in stackvs #1

#10GPT-5.6 Sol(OpenAI)

Score 66.8

SWE-bench77.6%Elo1,608Speed84 tok/sTTFT270 ms

Rank #10 in stackvs #1