Navigate
AI BENCHY
Advertise here

AI BENCHY Compare

Kwaipilot: KAT-Coder-Air V2.5 vs Qwen: Qwen3.5-35B-A3B

Summary

KAT-Coder-Air V2.5 (high) leads on average score with 6.2 vs 5.9. Qwen3.5-35B-A3B has the lower benchmark cost at $0.012 vs $0.064. Qwen3.5-35B-A3B is faster at 3.37s vs 11.02s, with pass rates of 46.0% vs 42.9%.

Recommended modelQwen3.5-35B-A3BIts score stays close to the best score here (5.9 vs 6.2), while costing about 5.7x less than KAT-Coder-Air V2.5 (high).

Last updated at: 2026-07-14

Metric KAT-Coder-Air V2.5 KAT-Coder-Air V2.5 high Release: 2026-07-14 Qwen3.5-35B-A3B Qwen3.5-35B-A3B none Release: 2026-02-24
Score 6.2 5.9
Rank #110 #127
Reliability 9.7 10.0
Consistency 7.6 8.9
Tests Correct
Attempt pass rate 46.0% 42.9%
Flaky tests 6 3
Total Runs 63 63
Cost per result 0.910 0.230
Total Cost $0.064 $0.012
Input Price $0.150 / 1M $0.140 / 1M
Output Price $0.600 / 1M $1.000 / 1M
Total Input Tokens 41,748 48,194
Output Tokens 3,894 4,343
Reasoning Tokens 91,773 0
Response Time (avg) 11.02s 3.37s
Response Time (max) 104.98s 47.43s
Response Time (total) 231.37s 70.75s

Generation showcase

Hamster playing table tennis

Prompt: Create a detailed SVG illustration of a hamster playing table tennis.

#110 KAT-Coder-Air V2.5

high
Invalid SVG
Cost
$0.000
Time
300.0s
Tokens
0 tok

#127 Qwen3.5-35B-A3B

none
Cost
$0.005
Time
28.4s
Tokens
4,518 tok

Top Models by Score

Score vs Total Cost

Response Time (avg)

Score vs Response Time (avg)

Total Output Tokens

Score vs Total Output Tokens

Category Breakdown

Coding Score Consistency Attempt pass rate Flaky tests Tests Correct Response Time (avg) Input Tokens Output Tokens Reasoning Tokens
KAT-Coder-Air V2.5 4.3 7.3 11.1% 1 39.07s 6,948 1,166 55,883
Qwen3.5-35B-A3B 5.5 10.0 33.3% 0 1.39s 7,808 571 0

Quick Compare

Switch Comparison Pair