Model benchmarks
KAT-Coder-V2.5-Dev.i1-IQ3_XXS local LLM performance
As of September 2026, KAT-Coder-V2.5-Dev.i1-IQ3_XXS runs at up to 86.9 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
86.9 tok/s
Average speed
85.7 tok/s
Avg PP
1198.1 tok/s
Min memory
15.6 GB
Max context
131,072 tokens
Avg output / run
9,227 tokens
Avg runtime / run
1m 54s
Avg quality
76.8
Benchmark runs
2
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 76.6 | 76.8 | 77.1 |
| Agent Workflow | 81.3 | 82.0 | 82.8 |
| Code Generation | 69.1 | 73.7 | 78.3 |
| Role Play & Narrative | 71.0 | 71.4 | 71.9 |
| Research & Analysis | 77.7 | 80.1 | 82.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination KAT-Coder-V2.5-Dev.i1-IQ3_XXS has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 5060 | llama.cpp | — | 86.9 tok/s | 85.7 tok/s | 15.6 GB | 131,072 tokens | 76.8 | 2 |
Benchmark runs
All 2 KAT-Coder-V2.5-Dev.i1-IQ3_XXS runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is KAT-Coder-V2.5-Dev.i1-IQ3_XXS good for coding?
- In our benchmarks, KAT-Coder-V2.5-Dev.i1-IQ3_XXS scores 73.7/100 for coding. It runs at about 85.7 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e is faster (~107.9 tok/s) and still scores well for coding (81.5/100).
- Is KAT-Coder-V2.5-Dev.i1-IQ3_XXS good for agentic (tool-using) tasks?
- In our benchmarks, KAT-Coder-V2.5-Dev.i1-IQ3_XXS scores 82.0/100 for agentic workflows. It runs at about 85.7 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e-MTP is faster (~123.7 tok/s) and still scores well for agentic workflows (84.4/100).
- How fast is KAT-Coder-V2.5-Dev.i1-IQ3_XXS for local inference?
- Across 2 community benchmark runs, KAT-Coder-V2.5-Dev.i1-IQ3_XXS reaches up to 86.9 tok/s and averages 85.7 tok/s, with the fastest results on NVIDIA GeForce RTX 5060.
- How much memory does KAT-Coder-V2.5-Dev.i1-IQ3_XXS need?
- The leanest observed configuration used about 15.6 GB of memory.
- Which tools have been used to run KAT-Coder-V2.5-Dev.i1-IQ3_XXS?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.