Model benchmarks

qwen3-coder-next:latest local LLM performance

As of May 2026, qwen3-coder-next:latest runs at up to 34.1 tok/s for local inference (best of 4 community benchmark runs across 1 GPU).

OllamaQ4_K_M

Model size

79.7B

Peak speed

34.1 tok/s

Average speed

31.4 tok/s

Min memory

50.3 GB

Max context

4,839 tokens

Best quality

94.5

Benchmark runs

4

GPUs tested

1

Performance by hardware and tool

Every hardware/tool/quantization combination qwen3-coder-next:latest has been benchmarked on, ranked by peak token generation speed. Last updated May 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
Apple M5 MaxOllamaQ4_K_M34.1 tok/s31.4 tok/s50.3 GB4,839 tokens74.94

Frequently asked questions

How fast is qwen3-coder-next:latest for local inference?
Across 4 community benchmark runs, qwen3-coder-next:latest reaches up to 34.1 tok/s and averages 31.4 tok/s, with the fastest results on Apple M5 Max.
How much memory does qwen3-coder-next:latest need?
The leanest observed configuration used about 50.3 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run qwen3-coder-next:latest?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.