Navigate
AI BENCHY
Advertise here

Kwaipilot: KAT-Coder-Pro V2.5 vs xAI: Grok 4.20

KAT-Coder-Pro V2.5 (high) leads on average score with 7.2 vs 7.1. KAT-Coder-Pro V2.5 (high) has the lower benchmark cost at $0.482 vs $0.777. KAT-Coder-Pro V2.5 (high) is faster at 20.83s vs 29.47s, with pass rates of 63.6% vs 63.6%.

Recommended modelKAT-Coder-Pro V2.5 (high)It has the best score here (7.2), while costing about 1.6x less than Grok 4.20 (medium).

Last updated at: 2026-07-17

Metric KAT-Coder-Pro V2.5 KAT-Coder-Pro V2.5 high Release: 2026-07-14 Grok 4.20 Grok 4.20 medium Release: 2026-03-31
Score 7.2 7.1
Rank #66 #72
Reliability 10.0 10.0
Consistency 7.8 8.5
Tests Correct
Attempt pass rate 63.6% 63.6%
Flaky tests 6 4
Total Runs 66 66
Cost per result 4.378 9.709
Total Cost $0.482 $0.777
Input Price $0.740 / 1M $1.250 / 1M
Output Price $2.960 / 1M $2.500 / 1M
Total Input Tokens 106,076 102,791
Output Tokens 9,071 5,363
Reasoning Tokens 127,093 253,977
Response Time (avg) 20.83s 29.47s
Response Time (max) 199.97s 199.66s
Response Time (total) 458.31s 648.35s

Hamster playing table tennis

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

#66 KAT-Coder-Pro V2.5

high
Cost
$0.009
Time
26.4s
Tokens
3,117 tok

#72 xAI: Grok 4.20

medium
Cost
$0.041
Time
110.3s
Tokens
16,336 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-Pro V2.5 6.4 7.9 44.4% 1 22.00s 7,893 422 20,461
Grok 4.20 6.3 6.6 55.6% 1 109.93s 8,307 268 103,150

Quick Compare

Switch Comparison Pair