Is Kwaipilot-KAT-Coder good for coding?

In our benchmarks, Kwaipilot-KAT-Coder scores 69.9/100 for coding. It runs at about 83.7 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).

Is Kwaipilot-KAT-Coder good for agentic (tool-using) tasks?

In our benchmarks, Kwaipilot-KAT-Coder scores 65.3/100 for agentic workflows. It runs at about 83.7 tok/s, so if you want more speed, Qwen3.8-Flash-Next-oQ3.5e-mtp is faster (~98.5 tok/s) and still scores well for agentic workflows (92.5/100).

How fast is Kwaipilot-KAT-Coder for local inference?

Across 6 community benchmark runs, Kwaipilot-KAT-Coder reaches up to 97.1 tok/s and averages 83.7 tok/s, with the fastest results on Tesla V100-PCIE-32GB.

How much memory does Kwaipilot-KAT-Coder need?

The leanest observed configuration used about 47.1 GB of memory.

Which tools have been used to run Kwaipilot-KAT-Coder?

Benchmarks were submitted using OpenAI-compatible, llama.cpp. Results are community-contributed and updated as new runs arrive.

Model benchmarks

Kwaipilot-KAT-Coder local LLM performance

As of November 2026, Kwaipilot-KAT-Coder runs at up to 97.1 tok/s for local inference (best of 6 community benchmark runs across 1 GPU).

OpenAI-compatiblellama.cpp

ShareReddit X

Model size

Unknown

Peak speed

97.1 tok/s

Average speed

83.7 tok/s

Avg PP

446.7 tok/s

Min memory

47.1 GB

Max context

32,768 tokens

Avg output / run

11,723 tokens

Avg runtime / run

Avg quality

76.0

Benchmark runs

GPUs tested

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 6 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	61.7	76.0	82.9
Agent Workflow	27.2	65.3	88.9
Code Generation	55.3	69.9	77.5
Role Play & Narrative	77.9	85.3	91.0
Research & Analysis	82.1	83.5	84.5

Performance by hardware and tool

Every hardware/tool/quantization combination Kwaipilot-KAT-Coder has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
Tesla V100-PCIE-32GB	llama.cpp	—	97.1 tok/s	85.0 tok/s	47.1 GB	32,768 tokens	72.8	4
Tesla V100-PCIE-32GB	OpenAI-compatible	—	81.2 tok/s	81.1 tok/s	47.3 GB	32,768 tokens	82.3	2

Benchmark runs

All 6 Kwaipilot-KAT-Coder runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is Kwaipilot-KAT-Coder good for coding?: In our benchmarks, Kwaipilot-KAT-Coder scores 69.9/100 for coding. It runs at about 83.7 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
Is Kwaipilot-KAT-Coder good for agentic (tool-using) tasks?: In our benchmarks, Kwaipilot-KAT-Coder scores 65.3/100 for agentic workflows. It runs at about 83.7 tok/s, so if you want more speed, Qwen3.8-Flash-Next-oQ3.5e-mtp is faster (~98.5 tok/s) and still scores well for agentic workflows (92.5/100).
How fast is Kwaipilot-KAT-Coder for local inference?: Across 6 community benchmark runs, Kwaipilot-KAT-Coder reaches up to 97.1 tok/s and averages 83.7 tok/s, with the fastest results on Tesla V100-PCIE-32GB.
How much memory does Kwaipilot-KAT-Coder need?: The leanest observed configuration used about 47.1 GB of memory.
Which tools have been used to run Kwaipilot-KAT-Coder?: Benchmarks were submitted using OpenAI-compatible, llama.cpp. Results are community-contributed and updated as new runs arrive.