Models

Loading community benchmarks…

← All models

Community benchmarks

Qwen3 4B Thinking 2507

Submitted benchmarks for Qwen3 4B Thinking 2507. Browse individual runs and their measurement settings, or open the report JSON for full benchmark details.

Compare chips Compare runtimes Add your benchmark

Submitted runs: 12
Chips tested: 6
Contributors: 2

Model identity

Grouped as Qwen/Qwen3-4B-Thinking-2507. The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.. Artifact verification confirms published model bytes, not that a community-submitted benchmark was honestly executed. Hover or tap a label for what it means.

This model group collects runs reported as basecompute/Qwen3-4B-Thinking-2507, Qwen3-4B-Thinking-2507-Q4, Qwen3-4B-Thinking-2507-Q8. Individual results keep the package name and format they were submitted with.

View on Hugging Face ↗

Published files benchmarked

basecompute/Qwen3-4B-Thinking-2507/Qwen3-4B-Thinking-2507-Q4.base (opens in a new tab)BaseRTQ4· 6 runs
basecompute/Qwen3-4B-Thinking-2507/Qwen3-4B-Thinking-2507-Q8.base (opens in a new tab)BaseRTQ8· 6 runs

Benchmark results

Peak throughput is the highest reported value and may come from different runs. Compare model, quantisation, backend, and token counts before drawing conclusions. Model verification identifies published artifact bytes; community benchmark execution is not remotely attested.

Individual runs for Qwen3 4B Thinking 2507. Throughput is in tokens per second; dates use your browser timezone. Hover a date for the exact local time.
Model / runtime / format	Chip / backend	Decode	Prefill	Peak process memory	Contributor

Qwen3-4B-Thinking-2507-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Pro Metal	78.4 tok/s TG128 @ 1 ctx	732 tok/s PP512	3,898.8 MiB	basecompute	20 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M3 Ultra Metal	155.6 tok/s TG128 @ 1 ctx	2,582 tok/s PP512	3,894.0 MiB	basecompute	20 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q8 BaseRTQ8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M3 Ultra Metal	118.2 tok/s TG128 @ 1 ctx	2,559 tok/s PP512	6,033.0 MiB	basecompute	20 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q8 BaseRTQ8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Pro Metal	56.3 tok/s TG128 @ 1 ctx	731 tok/s PP512	6,037.5 MiB	basecompute	20 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M1 Max Metal	93.2 tok/s TG128 @ 1 ctx	921 tok/s PP512	3,900.1 MiB	basecompute	20 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q8 BaseRTQ8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M1 Max Metal	71.5 tok/s TG128 @ 1 ctx	925 tok/s PP512	6,036.8 MiB	basecompute	20 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q8 BaseRTQ8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Max Metal	104.3 tok/s TG128 @ 1 ctx	1,714 tok/s PP512	6,036.0 MiB	basecompute	19 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M4 Max Metal	122.7 tok/s TG128 @ 1 ctx	1,721 tok/s PP512	3,897.1 MiB	basecompute	19 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q8 BaseRTQ8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	NVIDIA GB10 CUDA	54.7 tok/s TG128 @ 1 ctx	11,473 tok/s PP512	4,397.1 MiB	basecompute	19 Sep 2026	View Benchmark
Qwen3-4B-Thinking-2507-Q4 BaseRTQ4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	NVIDIA GB10 CUDA	73.2 tok/s TG128 @ 1 ctx	11,020 tok/s PP512	3,105.8 MiB	basecompute	19 Sep 2026	View Benchmark
basecompute/Qwen3-4B-Thinking-2507 BaseRTQ8· default-q8 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro Metal	64.7 tok/s TG128	3,414 tok/s PP512	6,047.3 MiB	arki05	11 Sep 2026	View Benchmark
basecompute/Qwen3-4B-Thinking-2507 BaseRTQ4· default-q4 The report's model SHA-256 matches a file published on Hugging Face, so the exact model bytes are known. This does not attest benchmark execution.	Apple M5 Pro Metal	91.6 tok/s TG128	3,363 tok/s PP512	3,908.7 MiB	arki05	11 Sep 2026	View Benchmark