Advertise here
#332

KAT Coder AIR V2.5

Kwaipilot Release: 2026-07-14 Tested on: 2026-10-01 02:59 kwaipilot/kat-coder-air-v2.5::low
~35B total (~3B active)MoEClosedEstimated
(high) (medium) (low) (default)

Summary

KAT Coder AIR V2.5 scores 5.0 on AI BENCHY and ranks #332. It has 9.9 reliability, a 42.0% pass rate, $0.046 total cost, and 10.21s average response time.

What makes KAT Coder AIR V2.5 unique: It stands out most in Instructions following, where it ranks #2, while Coding is its weakest area at #14.

Model facts

Researched on 2026-08-12

Estimated
Parameters
~35B total (~3B active)
Architecture
MoE
Availability
Closed
License
-

Best estimate from public evidence; the vendor did not disclose every value. Weights were announced but were not available for the exact checkpoint at research time.

Parameter source Availability source

Score

5.0

Consistency

7.8

Reliability

9.9

Total Cost (Current Price)

$0.046

Total Output Tokens

60,414

Total Input Tokens

61,094

Input Price

$0.150 / 1M

Output Price

$0.600 / 1M

Cache Read Price

N/A

Cache Write Price

N/A

Cache prices apply to input tokens. Reads reuse cached prompts; writes store them and can cost extra. Output tokens use the output price.

Tests Correct

Wrong Tests: 16

Attempt pass rate: 42.0%

Flaky tests

6

Flaky tests had mixed outcomes across runs (at least one pass and one fail).

Response Time (avg)

10.21s

Response Time (max): 86.23s

Response Time (total): 234.86s

Wrong answer: 7 Extra formatting: 4 API error: 3 Did not follow instructions: 2

Hamster playing table tennis

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

#332 KAT Coder AIR V2.5

low
Cost
$0.004
Time
29.9s
Tokens
6,008 tok

Run history

Tested on Score Reliability Tests Correct Total Cost Compare
2026-10-01 02:59 Re-test 5.0 9.9 $0.046 Current run
2026-07-16 23:32 New test added 5.4 9.9 $0.041 Compare
2026-07-14 11:05 Initial run 6.0 9.7 $0.030 Compare

This run used a different benchmark suite. Keep suite changes in mind when reading historical movement.

Run comparison

RunBenchmark coverageScoreConsistencyReliabilityTests CorrectFlaky testsTotal Output TokensTotal Input TokensTotal CostResponse Time (avg)
2026-10-01 02:59 · Current run69/69 attempts5.07.89.97/23660,41461,094$0.04610.21s
2026-07-16 23:32 · New test added66/66 attempts5.47.49.97/22752,89561,085$0.04110.09s
Difference—-0.4+0.40.00-1+7519+9+$0.005+119ms

Benchmark coverage differs: 69/69 attempts (Target: 3 repeats per test) versus 66/66 attempts (Target: 3 repeats per test). Totals and repeat-sensitive metrics are not directly comparable.

These two runs used different benchmark suites, so the deltas reflect both model changes and suite changes.

Price History

Historical pricing data for this model from OpenRouter.

Date Input Price Output Price Cache Read Price Cache Write Price
2026-08-14 12:51 $0.150 / 1M $0.600 / 1M N/A N/A

Charts

Choose the first model, then click a second model to open a side-by-side page.

Total Output Tokens

Score vs Total Output Tokens

Quick Compare

KAT Coder AIR V2.5lowvsGPT-4o-minidefaultKAT Coder AIR V2.5lowvsInkling SmallnoneFree AvailableKAT Coder AIR V2.5lowvsDots 3 Note PreviewdefaultFree AvailableKAT Coder AIR V2.5lowvsQwen3 Coder NextdefaultKAT Coder AIR V2.5lowvsMiniMax M2.7mediumKAT Coder AIR V2.5lowvsLaguna S 2.1highFree AvailableKAT Coder AIR V2.5lowvsMercury 2.5 PreviewlowKAT Coder AIR V2.5lowvsLing 2.6 1tdefaultKAT Coder AIR V2.5lowvsLaguna S 2.1mediumFree AvailableKAT Coder AIR V2.5lowvsMiMo-V2.5none

Category Breakdown

Category Score Consistency Tests Correct
Agentic 3.0 10.0
Anti-AI Tricks 7.3 5.8
Coding 3.5 7.3
Combined 6.4 5.8
Data parsing and extraction 6.5 10.0
Domain specific 3.0 10.0
General Intelligence 5.0 10.0
Instructions following 9.8 10.0
Puzzle Solving 3.1 4.6
Tool Calling 10.0 10.0
Trivia 3.0 10.0

Compared models