#79 MoonshotAI: Kimi K2.5
medium- Cost
- $0.030
- Time
- 58.6s
- Tokens
- 8,683 tok
Moonshot AI
Release: 2026-01-27
Tested on: 2026-05-22 00:12
moonshotai/kimi-k2.5::medium
1T total (32B active)MoEWeights available
(medium)
(none)
Summary
Kimi K2.5 scores 6.7 on AI BENCHY and ranks #79. It has 10.0 reliability, a 66.7% pass rate, $0.314 total cost, and 89.36s average response time.
What makes Kimi K2.5 unique: It stands out most in General Intelligence, where it ranks #4, while Coding is its weakest area at #17.
Model facts
Researched on 2026-08-12
- Parameters
- 1T total (32B active)
- Architecture
- MoE
- Availability
- Weights available
- License
- Modified MIT
6.7
Consistency
6.8
10.0
$0.314
Total Output Tokens
174,803
Total Input Tokens
0
Input Price
$0.400 / 1M
Output Price
$1.900 / 1M
Flaky tests
8
Flaky tests had mixed outcomes across runs (at least one pass and one fail).
Hamster playing table tennis
Prompt: Create a detailed SVG illustration of a hamster playing table tennis.
Run history
| Tested on | Score | Reliability | Tests Correct | Total Cost | Compare |
|---|---|---|---|---|---|
| 2026-08-14 02:00 Re-test | 7.0 | 10.0 | $0.607 ↑ | Compare | |
| 2026-07-16 22:17 New test added | 7.0 | 10.0 | $0.600 ↑ | Compare | |
| 2026-06-04 13:43 New test added | 6.8 | 10.0 | $0.328 ↓ | Compare | |
| 2026-05-22 00:12 Suite changed | 6.7 | 10.0 | $0.314 | Current run | |
| 2026-04-20 17:48 First recorded run | 7.0 | N/A | $0.220 | Compare |
This run used a different benchmark suite. Keep suite changes in mind when reading historical movement.
Run comparison
| Run | Benchmark coverage | Score | Consistency | Reliability | Tests Correct | Flaky tests | Total Output Tokens | Total Input Tokens | Total Cost | Response Time (avg) |
|---|---|---|---|---|---|---|---|---|---|---|
| 2026-05-22 00:12 · Suite changed | 60/60 attempts | 6.7 | 6.8 | 10.0 | 9/20 | 8 | 174,803 | 0 | $0.314 | 89.36s |
| 2026-04-20 17:48 · First recorded run | 54/54 attempts | 7.0 | 6.8 | N/A | 9/18 | 7 | 127,046 | 0 | $0.220 | 72.43s |
| Difference | — | -0.3 | 0.0 | 0 | +1 | +47757 | 0 | +$0.094 | +16930ms |
Benchmark coverage differs: 60/60 attempts (Target: 3 repeats per test) versus 54/54 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.
Charts
Choose the first model, then click a second model to open a side-by-side page.
Top Models by Score
Score vs Total Cost
Response Time (avg)
Score vs Response Time (avg)
Total Output Tokens
Score vs Total Output Tokens
Quick Compare
Kimi K2.5mediumvsSeed-2.0-MinimediumKimi K2.5mediumvsKAT-Coder-Pro V2.5mediumKimi K2.5mediumvsGrok 4.20mediumKimi K2.5mediumvsQwen3.7 FlashmediumKimi K2.5mediumvsMercury 2mediumKimi K2.5mediumvsSeed-2.0-CodehighKimi K2.5mediumvsGPT-5.5noneKimi K2.5mediumvsGemini 3.5 FlashnoneKimi K2.5mediumvsGLM 5.1mediumKimi K2.5mediumvsQwen3.6 Flashmedium
Category Breakdown
| Category | Score | Consistency | Tests Correct |
|---|---|---|---|
| Anti-AI Tricks | 7.3 | 5.8 | |
| Coding | 4.1 | 1.9 | |
| Combined | 10.0 | 10.0 | |
| Data parsing and extraction | 10.0 | 10.0 | |
| Domain specific | 3.5 | 4.4 | |
| General Intelligence | 6.5 | 3.4 | |
| Instructions following | 10.0 | 10.0 | |
| Puzzle Solving | 5.3 | 7.3 | |
| Tool Calling | 10.0 | 10.0 | |
| Trivia | 3.0 | 10.0 |