Skip to content
m.recipes
Find a configuration
Sign in
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

7 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 supportView original reportsMeasured alternatives for this taskTest your own GPU

Measured alternatives on other GPUs

Measured on different GPUs with the same or smaller listed capacity. These are references for choosing hardware, not performance or compatibility evidence for your selected GPU.

  1. Qwen3.8-27B-NVFP4

    RadixArk · 2 tested configurations · 1 GPU

    • RTX 5090 · 32 GBNVFP4 · 8K → 2K tokens
      Output throughput · includes prefill · 1 request
      70.8tok/s
      Observed GPU occupancy
      27.3GiB
      First token
      633 ms
      Use this recipe · 1 request ↗Evidence report and run availability ↗Same test on other GPUs
      Test details

      Total output tokens divided by the full request window, including prefill and queueing.

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

      Reports combined counts reports, not inference requests. Request successes and failures are listed in each report.

      Engine
      vllm
      Protocol
      wbench-fixed-http-v1
      Dataset
      synthetic-tokenizer-v1:repeats=2
      Workload
      8192in-2048out
      ConcurrencyMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
      C170.8229.01 s633 ms27.29405.51bad294c0 ↗
    • 1 other test configurations on this GPU
      • NVFP4 · 32K → 2K tokens
        Output throughput · includes prefill · 1 request
        62.5tok/s
        Observed GPU occupancy
        27.3GiB
        First token
        3.6 s
        Use this recipe · 1 request ↗Evidence report and run availability ↗Same test on other GPUs
        Test details

        Total output tokens divided by the full request window, including prefill and queueing.

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

        Reports combined counts reports, not inference requests. Request successes and failures are listed in each report.

        Engine
        vllm
        Protocol
        wbench-fixed-http-v1
        Dataset
        synthetic-tokenizer-v1:repeats=2
        Workload
        32768in-2048out
        ConcurrencyMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
        C162.4832.79 s3.6 s27.29433.31746bbc39 ↗

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

Quality by dataset

Scores from different datasets and sample scopes are not directly comparable.
Quality by dataset · 7 rows
ModelDatasetFormatEngineAverage scoreSample sizeRuns
RadixArk/Qwen3.8-27B-NVFP4hle-test-self-judge-22ed307-v1-n5-mt16384-t0NVFP4vllm20%5 / 2500 · Subset only
1 runs20% · 7bbadab5 ↗
RadixArk/Qwen3.8-27B-NVFP4simpleqa-verified-eval-self-judge-0dc97e0-v1-n1000-mt16384-t0NVFP4vllm6.4%1000 / 1000 · Full dataset
1 runs6.4% · 404eb015 ↗
RadixArk/Qwen3.8-27B-NVFP4simpleqa-verified-eval-self-judge-0dc97e0-v1-n5-mt16384-t0NVFP4vllm0%5 / 1000 · Subset only
1 runs0% · 328fd742 ↗
RadixArk/Qwen3.8-27B-NVFP4ifbench-test-official-loose-1c40f0c-v1-n300-mt16384-t0NVFP4vllm67.67%300 / 300 · Full dataset
1 runs67.67% · 02a43d63 ↗
RadixArk/Qwen3.8-27B-NVFP4ifbench-test-official-loose-1c40f0c-v1-n3-mt16384-t0NVFP4vllm66.67%3 / 300 · Subset only
1 runs66.67% · 20ef4b9f ↗
RadixArk/Qwen3.8-27B-NVFP4aime2026-test-zero-shot-strict-v4-n30-mt16384-t0NVFP4vllm63.33%30 / 30 · Full dataset
1 runs63.33% · 87acbd23 ↗
RadixArk/Qwen3.8-27B-NVFP4mmlu-pro-test-zero-shot-strict-v4-n3-mt16384-t0NVFP4vllm66.67%3 / 12032 · Subset only
1 runs66.67% · 1d2bdd31 ↗

Original reports

Every report for this task, including quality, diagnostic and legacy runs that do not enter comparisons.
Original reports · 7 rows
ModelTest taskFormatEngineWorkloadProtocolQuality scoreTest date
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmhle-test-self-judge-22ed307-v1-n5-mt16384-t0quality-v120%10/17/2026, 2:30:19 AM
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmsimpleqa-verified-eval-self-judge-0dc97e0-v1-n1000-mt16384-t0quality-v16.4%10/17/2026, 1:50:19 AM
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmsimpleqa-verified-eval-self-judge-0dc97e0-v1-n5-mt16384-t0quality-v10%10/15/2026, 9:23:17 PM
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmifbench-test-official-loose-1c40f0c-v1-n300-mt16384-t0quality-v167.67%10/15/2026, 8:57:08 PM
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmifbench-test-official-loose-1c40f0c-v1-n3-mt16384-t0quality-v166.67%10/15/2026, 9:27:35 AM
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmaime2026-test-zero-shot-strict-v4-n30-mt16384-t0quality-v163.33%10/15/2026, 9:07:41 AM
Qwen3.8-27B-NVFP4 ↗Text generationNVFP4vllmmmlu-pro-test-zero-shot-strict-v4-n3-mt16384-t0quality-v166.67%10/15/2026, 6:20:58 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, 9:56:52 PM