48 GB · 445 reports

RTX 4090D · 48 GB

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

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 · RTX 4090D · 48 GB

42 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.

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

Measurements without a comparable protocol · 110 rows

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

Measurements without a comparable protocol · 110 rows
ModelFormatEngineWorkloadConcurrencyThroughputGPU power (W)
Gemma-3-270m ↗BF16vLLMMixed1558.43 tok/s185.07
Gemma-3-270m ↗BF16vLLMMixed2861.15 tok/s164.61
Gemma-3-270m ↗BF16vLLMMixed41,559.22 tok/s178.04
Gemma-3-270m ↗BF16vLLMMixed82,479.12 tok/s161.51
Gemma-3-270m ↗BF16vLLM32K1532.47 tok/s189.12
Gemma-3-270m ↗BF16vLLM32K2827.48 tok/s173.12
Gemma-3-270m ↗BF16vLLM32K41,522.33 tok/s192.63
Gemma-3-270m ↗BF16vLLM32K81,458.27 tok/s157.46
Gemma-3-270m ↗BF16vLLM16K1546.4 tok/s186.78
Gemma-3-270m ↗BF16vLLM16K2856.9 tok/s169.84
Gemma-3-270m ↗BF16vLLM16K41,630.3 tok/s187.19
Gemma-3-270m ↗BF16vLLM16K81,552.35 tok/s149.86
Gemma-3-270m ↗BF16vLLM8K1553.45 tok/s184.86
Gemma-3-270m ↗BF16vLLM8K81,651.81 tok/s146.62
Gemma-3-270m ↗BF16vLLM8K325,116.35 tok/s192.84
Gemma-3-270m ↗BF16vLLM8K1288,584.2 tok/s252.28
Gemma-3-270m ↗BF16vLLM8K5128,452.13 tok/s267.55
Gemma-3-270m ↗BF16vLLM4K1566.15 tok/s185.97
Gemma-3-270m ↗BF16vLLM4K81,602.47 tok/s142.08
Gemma-3-270m ↗BF16vLLM4K325,431.31 tok/s184.86
Gemma-3-270m ↗BF16vLLM4K1288,967.4 tok/s240.86
Gemma-3-270m ↗BF16vLLM4K5129,211.88 tok/s261.53
Gemma-3-270m ↗BF16vLLM2K1566.41 tok/s184.31
Gemma-3-270m ↗BF16vLLM2K81,649.66 tok/s141.18
Gemma-3-270m ↗BF16vLLM2K325,509.22 tok/s179.5
Gemma-3-270m ↗BF16vLLM2K1289,516.01 tok/s234.75
Gemma-3-270m ↗BF16vLLM2K5129,785.93 tok/s260.17
Gemma-3-270m ↗BF16vLLM1K1566.96 tok/s184.81
Gemma-3-270m ↗BF16vLLM1K81,687.01 tok/s140.83
Gemma-3-270m ↗BF16vLLM1K325,586.19 tok/s178.08
Gemma-3-270m ↗BF16vLLM1K1289,836.61 tok/s232.35
Gemma-3-270m ↗BF16vLLM1K51210,434.61 tok/s262.5
Gemma-3-270m ↗BF16vLLM5121598.27 tok/s189.67
Gemma-3-270m ↗BF16vLLM51281,695.08 tok/s138.95
Gemma-3-270m ↗BF16vLLM512325,829.92 tok/s171.14
Gemma-3-270m ↗BF16vLLM51212810,407.53 tok/s218.17
Gemma-3-270m ↗BF16vLLM51251211,978.52 tok/s250.64
Gemma-3-270m ↗BF16vLLM2561640.38 tok/s195.23
Gemma-3-270m ↗BF16vLLM25681,917.24 tok/s142.39
Gemma-3-270m ↗BF16vLLM256326,067.09 tok/s161.77
Gemma-3-270m ↗BF16vLLM25612812,576.67 tok/s211.73
Gemma-3-270m ↗BF16vLLM25651214,801.51 tok/s230.89
Gemma-3-270m ↗BF16vLLM1281682.05 tok/s200.71
Gemma-3-270m ↗BF16vLLM12882,330.32 tok/s153.08
Gemma-3-270m ↗BF16vLLM128326,305 tok/s155.19
Gemma-3-270m ↗BF16vLLM12812814,741.29 tok/s209.43
Gemma-3-270m ↗BF16vLLM1281648.15 tok/s196.02
Gemma-3-270m ↗BF16vLLM12882,667.02 tok/s161.76
Gemma-3-270m ↗BF16vLLM128326,590.82 tok/s162.13
Gemma-3-270m ↗BF16vLLM12812814,844.36 tok/s218.56
Gemma-3-270m ↗BF16vLLM12851218,775.66 tok/s242.81
Gemma-3-270m ↗BF16vLLMMixed1327.11 tok/s173.29
Gemma-3-270m ↗BF16vLLMMixed2358.96 tok/s184.1
Gemma-3-270m ↗BF16vLLMMixed4431.84 tok/s235.46
Gemma-3-270m ↗BF16vLLMMixed8331.25 tok/s228.1
Gemma-3-270m ↗BF16vLLM4K1447.36 tok/s174.14
Gemma-3-270m ↗BF16vLLM4K81,552.03 tok/s149.96
Gemma-3-270m ↗BF16vLLM4K324,736.15 tok/s188.5
Gemma-3-270m ↗BF16vLLM4K1287,607.28 tok/s243.69
Gemma-3-270m ↗BF16vLLM4K5125,229.81 tok/s252.58
Gemma-3-270m ↗BF16vLLM2K1562.52 tok/s185.58
Gemma-3-270m ↗BF16vLLM2K325,235.34 tok/s181.53
Gemma-3-270m ↗BF16vLLM2K1288,913.21 tok/s243.45
Gemma-3-270m ↗BF16vLLM2K5127,885.21 tok/s266.09
Gemma-3-270m ↗BF16vLLM32K1536.84 tok/s190.17
Gemma-3-270m ↗BF16vLLM16K1260.85 tok/s152.74
Gemma-3-270m ↗BF16vLLM1K81,599.27 tok/s149.04
Gemma-3-270m ↗BF16vLLM1K325,070.66 tok/s185.03
Gemma-3-270m ↗BF16vLLM1K1289,108.9 tok/s248.31
Gemma-3-270m ↗BF16vLLM512137.2 tok/s269.12
Gemma-3-270m ↗BF16vLLM51281,339.85 tok/s185.83
Gemma-3-270m ↗BF16vLLM512325,806.54 tok/s172.88
Gemma-3-270m ↗BF16vLLM51212810,299.83 tok/s228.27
Gemma-3-270m ↗BF16vLLM5125128,939.01 tok/s261.82
Gemma-3-270m ↗BF16vLLM2K1436.43 tok/s170.53
Gemma-3-270m ↗BF16vLLM2K81,445.3 tok/s164.85
Gemma-3-270m ↗BF16vLLM2K325,480.34 tok/s181.78
Gemma-3-270m ↗BF16vLLM2K1286,609.37 tok/s263.43
Gemma-3-270m ↗BF16vLLM2K5129,695.96 tok/s263.99
Gemma-3-270m ↗BF16vLLM25681,632.06 tok/s141.7
Gemma-3-270m ↗BF16vLLM256325,676.02 tok/s176.12
Gemma-3-270m ↗BF16vLLM2561286,102.07 tok/s253.16
Gemma-3-270m ↗BF16vLLM25651214,312.31 tok/s236.52
Gemma-3-270m ↗BF16vLLM1K1565.84 tok/s184.72
Gemma-3-270m ↗BF16vLLM1K81,665.74 tok/s141.35
Gemma-3-270m ↗BF16vLLM1K325,519.73 tok/s177.48
Gemma-3-270m ↗BF16vLLM1K1289,781.99 tok/s237.32
Gemma-3-270m ↗BF16vLLM1K51210,115.03 tok/s263.25
Gemma-3-270m ↗BF16vLLM5121608.88 tok/s191.83
Gemma-3-270m ↗BF16vLLM51281,713.17 tok/s140.12
Gemma-3-270m ↗BF16vLLM512322,851.12 tok/s205.79
Gemma-3-270m ↗BF16vLLM51212810,816.18 tok/s219.66
Gemma-3-270m ↗BF16vLLM51251212,144.95 tok/s252.42
Gemma-3-270m ↗BF16vLLM1281267.49 tok/s146.51
Gemma-3-270m ↗BF16vLLM12881,699.55 tok/s158.78
Gemma-3-270m ↗BF16vLLM128326,225.67 tok/s157.92
Gemma-3-270m ↗BF16vLLM12812812,831.27 tok/s199.11
Gemma-3-270m ↗BF16vLLM12851214,492.99 tok/s231.59
Gemma-3-270m ↗BF16vLLM25681,747.17 tok/s143.73
Gemma-3-270m ↗BF16vLLM256326,119.2 tok/s169.44
Gemma-3-270m ↗BF16vLLM25612812,219.3 tok/s205.72
Gemma-3-270m ↗BF16vLLM25651214,354.97 tok/s230.05
Gemma-3-270m ↗BF16vLLM1281681.68 tok/s200.79
Gemma-3-270m ↗BF16vLLM12881,846.76 tok/s138.58
Gemma-3-270m ↗BF16vLLM128326,287.08 tok/s156.81
Gemma-3-270m ↗BF16vLLM1281692.79 tok/s202.09
Gemma-3-270m ↗BF16vLLM12882,225.08 tok/s146.97
Gemma-3-270m ↗BF16vLLM128326,494.38 tok/s156.56
Gemma-3-270m ↗BF16vLLM12812815,588.69 tok/s212.36
Gemma-3-270m ↗BF16vLLM12851218,723.26 tok/s227.31

Original reports

Every report for this task, including quality, diagnostic and legacy runs that do not enter comparisons.
Original reports · 42 rows
ModelTest taskFormatEngineWorkloadProtocolQuality scoreTest date
gemma-3-270m ↗Text generationNativevLLM2KMethod unknown—1/25/2026, 4:47:46 AM
gemma-3-270m ↗Text generationNativevLLM1KMethod unknown—1/25/2026, 4:33:01 AM
gemma-3-270m ↗Text generationNativevLLM512Method unknown—1/25/2026, 4:23:12 AM
gemma-3-270m ↗Text generationNativevLLM256Method unknown—1/25/2026, 4:14:45 AM
gemma-3-270m ↗Text generationNativevLLM128Method unknown—1/25/2026, 4:07:02 AM
Gemma-3-270m ↗Text generationBF16vLLMMixedMethod unknown—3/6/2026, 10:54:28 PM
Gemma-3-270m ↗Text generationBF16vLLM128KMethod unknown—3/6/2026, 10:51:17 PM
Gemma-3-270m ↗Text generationBF16vLLM64KMethod unknown—3/6/2026, 10:47:54 PM
Gemma-3-270m ↗Text generationBF16vLLM32KMethod unknown—3/6/2026, 10:43:08 PM
Gemma-3-270m ↗Text generationBF16vLLM16KMethod unknown—3/6/2026, 10:39:20 PM
Gemma-3-270m ↗Text generationBF16vLLM8KMethod unknown—3/6/2026, 10:29:22 PM
Gemma-3-270m ↗Text generationBF16vLLM4KMethod unknown—3/6/2026, 10:21:36 PM
Gemma-3-270m ↗Text generationBF16vLLM2KMethod unknown—3/6/2026, 10:15:08 PM
Gemma-3-270m ↗Text generationBF16vLLM1KMethod unknown—3/6/2026, 10:08:36 PM
Gemma-3-270m ↗Text generationBF16vLLM512Method unknown—3/6/2026, 10:01:36 PM
Gemma-3-270m ↗Text generationBF16vLLM256Method unknown—3/6/2026, 9:50:42 PM
Gemma-3-270m ↗Text generationBF16vLLM128Method unknown—3/6/2026, 9:25:21 PM
Gemma-3-270m ↗Text generationBF16vLLM128Method unknown—3/6/2026, 9:06:52 PM
Gemma-3-270m ↗Text generationBF16vLLM8KMethod unknown—3/6/2026, 9:01:54 PM
Gemma-3-270m ↗Text generationBF16vLLMMixedMethod unknown—3/6/2026, 8:55:56 PM
Gemma-3-270m ↗Text generationBF16vLLM4KMethod unknown—3/6/2026, 8:55:28 PM
Gemma-3-270m ↗Text generationBF16vLLM128KMethod unknown—3/6/2026, 8:51:38 PM
Gemma-3-270m ↗Text generationBF16vLLM64KMethod unknown—3/6/2026, 8:48:24 PM
Gemma-3-270m ↗Text generationBF16vLLM2KMethod unknown—3/6/2026, 8:46:56 PM
Gemma-3-270m ↗Text generationBF16vLLM32KMethod unknown—3/6/2026, 8:45:10 PM
Gemma-3-270m ↗Text generationBF16vLLM16KMethod unknown—3/6/2026, 8:40:32 PM
Gemma-3-270m ↗Text generationBF16vLLM1KMethod unknown—3/6/2026, 8:39:50 PM
Gemma-3-270m ↗Text generationBF16vLLM8KMethod unknown—3/6/2026, 8:34:40 PM
Gemma-3-270m ↗Text generationBF16vLLM512Method unknown—3/6/2026, 8:31:13 PM
Gemma-3-270m ↗Text generationBF16vLLM4KMethod unknown—3/6/2026, 8:26:23 PM
Gemma-3-270m ↗Text generationBF16vLLM2KMethod unknown—3/6/2026, 8:17:37 PM
Gemma-3-270m ↗Text generationBF16vLLM256Method unknown—3/6/2026, 8:16:37 PM
Gemma-3-270m ↗Text generationBF16vLLM1KMethod unknown—3/6/2026, 8:09:47 PM
Gemma-3-270m ↗Text generationBF16vLLM512Method unknown—3/6/2026, 8:02:55 PM
Gemma-3-270m ↗Text generationBF16vLLM128Method unknown—3/6/2026, 7:53:02 PM
Gemma-3-270m ↗Text generationBF16vLLM256Method unknown—3/6/2026, 7:52:24 PM
Gemma-3-270m ↗Text generationBF16vLLM128Method unknown—3/6/2026, 7:43:39 PM
Gemma-3-270m ↗Text generationBF16vLLM128Method unknown—3/6/2026, 4:29:01 PM
Gemma-3-270m ↗Text generationBF16vLLM1KMethod unknown—1/25/2026, 6:03:52 AM
Gemma-3-270m ↗Text generationBF16vLLM512Method unknown—1/25/2026, 5:25:46 AM
Gemma-3-270m ↗Text generationBF16vLLM256Method unknown—1/25/2026, 5:16:21 AM
Gemma-3-270m ↗Text generationBF16vLLM128Method unknown—1/25/2026, 5:07:39 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