Model benchmarks

qwen3-8b-fast local LLM performance

As of November 2026, qwen3-8b-fast runs at up to 114.9 tok/s for local inference (best of 4 community benchmark runs across 1 GPU).

OpenAI-compatiblellama.cpp
ShareRedditX

Model size

8B

Peak speed

114.9 tok/s

Average speed

114.4 tok/s

Avg PP

3294.5 tok/s

Min memory

8.5 GB

Max context

65,536 tokens

Avg output / run

5,148 tokens

Avg runtime / run

57s

Avg quality

62.2

Benchmark runs

4

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 4 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall57.962.266.1
Agent Workflow69.574.780.1
Code Generation27.141.554.4
Role Play & Narrative69.275.284.0
Research & Analysis51.657.562.7

Performance by hardware and tool

Every hardware/tool/quantization combination qwen3-8b-fast has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
NVIDIA GeForce RTX 3080 Tillama.cpp—114.9 tok/s114.5 tok/s8.5 GB65,536 tokens62.03
NVIDIA GeForce RTX 3080 TiOpenAI-compatible—114.0 tok/s114.0 tok/s8.5 GB40,960 tokens63.01

Benchmark runs

All 4 qwen3-8b-fast runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is qwen3-8b-fast good for agentic (tool-using) tasks?
In our benchmarks, qwen3-8b-fast scores 74.7/100 for agentic workflows. It runs at about 114.4 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e-MTP is faster (~151.8 tok/s) and still scores well for agentic workflows (83.0/100).
How fast is qwen3-8b-fast for local inference?
Across 4 community benchmark runs, qwen3-8b-fast reaches up to 114.9 tok/s and averages 114.4 tok/s, with the fastest results on NVIDIA GeForce RTX 3080 Ti.
How much memory does qwen3-8b-fast need?
The leanest observed configuration used about 8.5 GB of memory.
Which tools have been used to run qwen3-8b-fast?
Benchmarks were submitted using OpenAI-compatible, llama.cpp. Results are community-contributed and updated as new runs arrive.