10 GB × 7 · 3 reports

7 × RTX 3080 · 10 GB

Qualified results first; original reports and full tables stay below.

Text generation · Qualified results

Speeds compare only within one exact test. Recipe availability is checked on each report against the published recipe executor.

No qualified results yet · Text generation · 7 × RTX 3080 · 10 GB

3 original reports are kept as evidence but do not qualify for comparison; each report states why. This does not show that the model or GPU cannot run it.

Check model launch support View original reports Measured alternatives for this task Test your own GPU

Occupancy includes engine buffers and reserved KV cache. It is not the minimum VRAM required by the model.

Measurements without a comparable protocol · 26 rows

These measured rates remain visible, but are excluded from comparison curves because the protocol is unknown.

Measurements without a comparable protocol · 26 rows
Model	Format	Engine	Workload	Concurrency	Throughput	GPU power (W)
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	1	74.65 tok/s	942.13
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	8	556.26 tok/s	833.64
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	16	1,108.16 tok/s	824.76
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	32	1,719.81 tok/s	987.62
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	64	3,720.19 tok/s	1,126.08
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	128	7,112.87 tok/s	1,300.14
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	256	13,148.15 tok/s	1,515.6
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	512	17,153.86 tok/s	1,915.19
Qwen3-0.6B ↗	BF16	vLLM-DP7	2K	1024	16,840.27 tok/s	1,968.44
Qwen3-1.7B ↗	BF16	vLLM	2K	1	70.9 tok/s	1,307.2
Qwen3-1.7B ↗	BF16	vLLM	2K	8	549.67 tok/s	1,413.1
Qwen3-1.7B ↗	BF16	vLLM	2K	16	1,095.47 tok/s	1,456.66
Qwen3-1.7B ↗	BF16	vLLM	2K	32	2,153.82 tok/s	1,532.92
Qwen3-1.7B ↗	BF16	vLLM	2K	64	3,681.68 tok/s	1,508.45
Qwen3-1.7B ↗	BF16	vLLM	2K	128	7,084.99 tok/s	1,632.15
Qwen3-1.7B ↗	BF16	vLLM	2K	256	11,357.29 tok/s	1,857.15
Qwen3-1.7B ↗	BF16	vLLM	2K	512	10,798.07 tok/s	1,999.1
Qwen3-1.7B ↗	BF16	vLLM	2K	1024	10,247.58 tok/s	1,928.95
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	1	54.38 tok/s	1,079.17
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	8	417.62 tok/s	1,232
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	16	833.77 tok/s	1,288.6
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	32	1,655.47 tok/s	1,367.27
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	64	2,837.46 tok/s	1,419.55
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	128	4,991.87 tok/s	1,651.09
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	256	8,732.6 tok/s	1,957.76
Qwen3-4B-AWQ ↗	AWQ	vLLM	2K	512	8,049.02 tok/s	2,062.1

Original reports

Every report for this task, including quality, diagnostic and legacy runs that do not enter comparisons.

Original reports · 3 rows
Model	Test task	Format	Engine	Workload	Protocol	Quality score	Test date
Qwen3-1.7B ↗	Text generation	BF16	vLLM	2K	Method unknown	—	1/15/2026, 9:52:50 AM
Qwen3-0.6B ↗	Text generation	BF16	vLLM-DP7	2K	Method unknown	—	1/15/2026, 9:08:58 AM
Qwen3-4B-AWQ ↗	Text generation	AWQ	vLLM	2K	Method unknown	—	1/15/2026, 10:10:06 AM

Running benchmarks

0Updates automatically

No benchmarks are currently reported as running.

Methodology

Average of error-free runs for the selected recorded configuration. Missing methods and fallback runs are not ranked. Performance index = average / fastest matching hardware × 100. Value index = performance per purchase cost / best comparable value × 100. Quality is separate by dataset and protocol. Prices are dated seller listings, not completed transactions or market averages.

Updated 11/5/2026, 11:02:27 PM