KAT-Coder-V2.5-Dev model card (Hugging Face) Jul 23, 2026
KAT-Coder-V2.5-Dev
KAT-Coder-V2.5-Dev places in the top three on PinchBench.
7
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Kuaishou Jul 2026 KAT-Coder-V2.5-Dev
Open weights Small
142.0
#163 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- SciCode 44.2% #22 /49
- SWE-bench Multilingual 63% #56 /73
- PinchBench 93.4% #2 /14
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SciCode | 44.2% | #22 of 49 | Lab model card | |
| SWE-bench Verified | 69.4% | #83 of 155 | Lab model card | |
| SWE-bench Multilingual | 63% | #56 of 73 | Lab model card | |
| SWE-Bench Pro | 46% | #76 of 81 | Lab model card | |
| KAT Code Bench | 46.2% | #2 of 2 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PinchBench | 93.4% | #2 of 14 | Lab model card | |
| Terminal-Bench 2.1 | 41% | #53 of 56 | Lab model card |
Its lab's numbers put it at 142.0, number 163 of 404
The Self-Reported Index fits one score to the 6 capability results Kuaishou reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 138.9 to 145.1. How we compute it.
- PinchBench 93.4
- SWE-bench Verified 69.4
- SWE-bench Multilingual 63
- SWE-Bench Pro 46
- SciCode 44.2
- Terminal-Bench 2.1 41
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.