Skip to content
m.recipes
Find a configuration
Sign in
5 models with qualified results

Models

Measured configurations by model. Open a model to compare GPUs on one exact test.

Get your startup command

Choose your operating system and enter a model repository. Launch support is checked separately from performance results.

Reset

All tasks · L4 · 24 GB

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

  1. Qwen3-4B-AWQ

    Qwen · 2 tested configurations

    • AWQ · 8K → 2K tokens · No cache reuse
      Output throughput · includes prefill · 1 request
      55.3tok/s
      Best measured · 4 concurrent requests
      109tok/s
      Observed GPU occupancy
      19.5GiB
      First token
      1.6 s
      Use this recipe · 4 concurrent requests ↗Evidence report and run availability ↗Same test on other GPUs
      Test details · 2 concurrency levels

      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 · vLLM 0.29.0
      Protocol
      wbench-fixed-http-v3
      Dataset
      migration-analysis-v1:sha256=68590c248e43f32c04dfd8062464fd1e4856917795fe307225e5c6112a54fe94:cache=isolated:repeats=3
      Workload
      8192in-2048out
      ConcurrencyMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
      C155.337.17 s1.6 s19.5471.919301a9ba ↗
      C4109.4378.88 s3.1 s19.5571.919301a9ba ↗
    • 1 other test configurations on this GPU
      • AWQ · 8K → 2K tokens · Prefix cache reuse
        Output throughput · includes prefill · 1 request
        57.4tok/s
        Best measured · 4 concurrent requests
        175tok/s
        Observed GPU occupancy
        19.5GiB
        First token
        60 ms
        Use this recipe · 4 concurrent requests ↗Evidence report and run availability ↗Same test on other GPUs
        Test details · 2 concurrency levels

        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 · vLLM 0.29.0
        Protocol
        wbench-fixed-http-v3
        Dataset
        migration-analysis-v1:sha256=68590c248e43f32c04dfd8062464fd1e4856917795fe307225e5c6112a54fe94:cache=prefix-reuse:repeats=3
        Workload
        8192in-2048out
        ConcurrencyMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
        C157.3935.7 s60 ms19.5571.817e2cbfda ↗
        C4175.3946.75 s114 ms19.5571.917e2cbfda ↗

  2. embeddinggemma-2

    unsloth · 2 tested configurations

    • nf4 · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s
      Throughput · Batch 1
      175vectors/sec
      Best measured · Batch 16
      1,650vectors/sec
      Observed GPU occupancy
      0.9GiB
      P95 latency
      6 ms
      Use this recipe · Batch 16 ↗Evidence report and run availability ↗Same test on other GPUs
      Test details · 3 batch sizes

      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
      sentence-transformers · PyTorch 2.12.1+cu126
      Protocol
      native-forward-v1
      Dataset
      embedding-mteb-tasks-v1
      Workload
      embedding
      Batch sizeMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
      B1174.956 ms—0.8554.8529b19870 ↗
      B4572.67 ms—0.8558.4529b19870 ↗
      B161,649.6210 ms—0.8560.7529b19870 ↗
    • 1 other test configurations on this GPU
      • fp16 · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s
        Throughput · Batch 1
        177vectors/sec
        Best measured · Batch 16
        2,036vectors/sec
        Observed GPU occupancy
        1GiB
        P95 latency
        6 ms
        Use this recipe · Batch 16 ↗Evidence report and run availability ↗Same test on other GPUs
        Test details · 3 batch sizes

        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
        sentence-transformers · PyTorch 2.12.1+cu126
        Protocol
        native-forward-v1
        Dataset
        embedding-mteb-tasks-v1
        Workload
        embedding
        Batch sizeMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
        B1177.216 ms—0.9950.54d74f69cc ↗
        B4648.086 ms—0.9957.14d74f69cc ↗
        B162,035.678 ms—0.9958.14d74f69cc ↗

  3. embeddinggemma-2

    google · 2 tested configurations

    • nf4 · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s
      Throughput · Batch 1
      175vectors/sec
      Best measured · Batch 16
      1,639vectors/sec
      Observed GPU occupancy
      0.9GiB
      P95 latency
      6 ms
      Use this recipe · Batch 16 ↗Evidence report and run availability ↗Same test on other GPUs
      Test details · 3 batch sizes

      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
      sentence-transformers · PyTorch 2.12.1+cu126
      Protocol
      native-forward-v1
      Dataset
      embedding-mteb-tasks-v1
      Workload
      embedding
      Batch sizeMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
      B1174.886 ms—0.8553.167ad6a821 ↗
      B4571.037 ms—0.855767ad6a821 ↗
      B161,639.0810 ms—0.8559.867ad6a821 ↗
    • 1 other test configurations on this GPU
      • fp16 · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s
        Throughput · Batch 1
        177vectors/sec
        Best measured · Batch 16
        2,039vectors/sec
        Observed GPU occupancy
        1GiB
        P95 latency
        6 ms
        Use this recipe · Batch 16 ↗Evidence report and run availability ↗Same test on other GPUs
        Test details · 3 batch sizes

        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
        sentence-transformers · PyTorch 2.12.1+cu126
        Protocol
        native-forward-v1
        Dataset
        embedding-mteb-tasks-v1
        Workload
        embedding
        Batch sizeMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
        B1177.466 ms—0.9949.86e0700e04 ↗
        B4650.166 ms—0.9956.96e0700e04 ↗
        B162,039.178 ms—0.9958.46e0700e04 ↗

  4. granite-embedding-97m-multilingual-r2

    ibm-granite · 1 tested configuration

    • nf4 · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s
      Throughput · Batch 1
      792vectors/sec
      Best measured · Batch 16
      8,881vectors/sec
      Observed GPU occupancy
      0.5GiB
      P95 latency
      1 ms
      Use this recipe · Batch 16 ↗Evidence report and run availability ↗Same test on other GPUs
      Test details · 3 batch sizes

      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
      sentence-transformers · PyTorch 2.12.1+cu126
      Protocol
      native-forward-v1
      Dataset
      embedding-mteb-tasks-v1
      Workload
      embedding
      Batch sizeMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
      B1791.851 ms—0.5137.41d396c54b ↗
      B42,816.451 ms—0.5137.91d396c54b ↗
      B168,880.862 ms—0.5138.91d396c54b ↗

  5. Qwen3.8-27B-UD-Q4_K_M.gguf

    unsloth/Qwen3.8-27B-GGUF · 2 tested configurations

    • Q4_K_M · 8K → 2K tokens
      Output throughput · includes prefill · 1 request
      14.1tok/s
      Observed GPU occupancy
      17GiB
      First token
      279 ms
      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
      llamacpp
      Protocol
      wbench-fixed-http-v1
      Dataset
      synthetic-tokenizer-v1:repeats=3
      Workload
      8192in-2048out
      ConcurrencyMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
      C114.09145.5 s279 ms17.0571.914bc32a6c ↗
    • 1 other test configurations on this GPU
      • Q4_K_M · 32K → 2K tokens
        Output throughput · includes prefill · 1 request
        12.9tok/s
        Observed GPU occupancy
        17GiB
        First token
        371 ms
        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
        llamacpp
        Protocol
        wbench-fixed-http-v1
        Dataset
        synthetic-tokenizer-v1:repeats=3
        Workload
        32768in-2048out
        ConcurrencyMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
        C112.94158.32 s371 ms17.0571.81d8a4b66b ↗
Sortable tables by exact test condition · 21 rows
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 1 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu126174.956 ms—0.8554.8529b19870 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 1 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBfp16sentence-transformersPyTorch 2.12.1+cu126177.216 ms—0.9950.54d74f69cc ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 1 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu126174.886 ms—0.8553.167ad6a821 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 1 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBfp16sentence-transformersPyTorch 2.12.1+cu126177.466 ms—0.9949.86e0700e04 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 1 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
granite-embedding-97m-multilingual-r2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu126791.851 ms—0.5137.41d396c54b ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 4 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu126572.67 ms—0.8558.4529b19870 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 4 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBfp16sentence-transformersPyTorch 2.12.1+cu126648.086 ms—0.9957.14d74f69cc ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 4 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu126571.037 ms—0.855767ad6a821 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 4 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBfp16sentence-transformersPyTorch 2.12.1+cu126650.166 ms—0.9956.96e0700e04 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 4 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
granite-embedding-97m-multilingual-r2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu1262,816.451 ms—0.5137.91d396c54b ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 16 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu1261,649.6210 ms—0.8560.7529b19870 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 16 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBfp16sentence-transformersPyTorch 2.12.1+cu1262,035.678 ms—0.9958.14d74f69cc ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 16 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu1261,639.0810 ms—0.8559.867ad6a821 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 16 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
embeddinggemma-2L4 · 24 GBfp16sentence-transformersPyTorch 2.12.1+cu1262,039.178 ms—0.9958.46e0700e04 ↗
Embedding · Text → Vectors · embedding-mteb-tasks-v1 · Test duration 90 s · Batch 16 · vectors/sec · 1 rows
ModelHardwareFormatEngineMean throughput (vectors/sec)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
granite-embedding-97m-multilingual-r2L4 · 24 GBnf4sentence-transformersPyTorch 2.12.1+cu1268,880.862 ms—0.5138.91d396c54b ↗
Text generation · 8K → 2K tokens · 1 request · tok/s · 1 rows
ModelHardwareFormatEngineMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
Qwen3.8-27B-UD-Q4_K_M.ggufL4 · 24 GBQ4_K_Mllamacpp14.09145.5 s279 ms17.0571.914bc32a6c ↗
Text generation · 8K → 2K tokens · No cache reuse · 1 request · tok/s · 1 rows
ModelHardwareFormatEngineMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
Qwen3-4B-AWQL4 · 24 GBAWQvllmvLLM 0.29.055.337.17 s1.6 s19.5471.919301a9ba ↗
Text generation · 8K → 2K tokens · No cache reuse · 4 concurrent requests · tok/s · 1 rows
ModelHardwareFormatEngineMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
Qwen3-4B-AWQL4 · 24 GBAWQvllmvLLM 0.29.0109.4378.88 s3.1 s19.5571.919301a9ba ↗
Text generation · 8K → 2K tokens · Prefix cache reuse · 1 request · tok/s · 1 rows
ModelHardwareFormatEngineMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
Qwen3-4B-AWQL4 · 24 GBAWQvllmvLLM 0.29.057.3935.7 s60 ms19.5571.817e2cbfda ↗
Text generation · 8K → 2K tokens · Prefix cache reuse · 4 concurrent requests · tok/s · 1 rows
ModelHardwareFormatEngineMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
Qwen3-4B-AWQL4 · 24 GBAWQvllmvLLM 0.29.0175.3946.75 s114 ms19.5571.917e2cbfda ↗
Text generation · 32K → 2K tokens · 1 request · tok/s · 1 rows
ModelHardwareFormatEngineMean throughput (tok/s)P95TTFTObserved occupancy (GiB)GPU power (W)Reports combinedView report
Qwen3.8-27B-UD-Q4_K_M.ggufL4 · 24 GBQ4_K_Mllamacpp12.94158.32 s371 ms17.0571.81d8a4b66b ↗
Models with reports but no qualified results · 28

Their original reports remain available on each model page. They did not qualify for comparison; each report states why.

Models with reports but no qualified results · 28 rows
ModelTest task
Llama-3.1-8B-Instruct-FP8nvidia/Llama-3.1-8B-Instruct-FP8Text generation
Ministral-3-3B-Instruct-2512mistralai/Ministral-3-3B-Instruct-2512Text generation
Ministral-3-8B-Instruct-2512mistralai/Ministral-3-8B-Instruct-2512Text generation
Ministral-3-14B-Instruct-2512mistralai/Ministral-3-14B-Instruct-2512Text generation
Ministral-3-14B-Instruct-2512-FP8mistralai/Ministral-3-14B-Instruct-2512-FP8Text generation
NVIDIA-Nemotron-3-Nano-4B-FP8nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8Text generation
NVIDIA-Nemotron-Nano-9B-v2-FP8nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8Text generation
NVIDIA-Nemotron-Nano-12B-v2-VL-FP8nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8Text generation
Qwen2.5-0.5Bqwen2.5:0.5bText generation
qwen2.5-0.5b-instruct-q4_k_m.ggufQwen/Qwen2.5-0.5B-Instruct-GGUF/qwen2.5-0.5b-instruct-q4_k_m.ggufText generation
qwen2.5-1.5b-instruct-q4_k_m.ggufQwen/Qwen2.5-1.5B-Instruct-GGUF/qwen2.5-1.5b-instruct-q4_k_m.ggufText generation
Qwen2.5-3Bqwen2.5:3bText generation
qwen2.5-3b-instruct-q4_k_m.ggufQwen/Qwen2.5-3B-Instruct-GGUF/qwen2.5-3b-instruct-q4_k_m.ggufText generation
Qwen3-0.6BQwen/Qwen3-0.6BText generation
Qwen3-0.6B-FP8Qwen/Qwen3-0.6B-FP8Text generation
Qwen3-0.6B-GPTQ-Int8Qwen/Qwen3-0.6B-GPTQ-Int8Text generation
Qwen3-1.7B-FP8Qwen/Qwen3-1.7B-FP8Text generation
Qwen3-4B-FP8Qwen/Qwen3-4B-FP8Text generation
Qwen3-4B-Instruct-2507-FP8Qwen/Qwen3-4B-Instruct-2507-FP8Text generation
Qwen3-4B-Thinking-2507-FP8Qwen/Qwen3-4B-Thinking-2507-FP8Text generation
Qwen3-8B-AWQQwen/Qwen3-8B-AWQText generation
Qwen3-8B-FP8Qwen/Qwen3-8B-FP8Text generation
Qwen3-14B-AWQQwen/Qwen3-14B-AWQText generation
Qwen3-14B-FP8Qwen/Qwen3-14B-FP8Text generation
Qwen3-VL-2B-Instruct-FP8Qwen/Qwen3-VL-2B-Instruct-FP8Text generation
Qwen3-VL-4B-Instruct-FP8Qwen/Qwen3-VL-4B-Instruct-FP8Text generation
Qwen3-VL-8B-Instruct-FP8Qwen/Qwen3-VL-8B-Instruct-FP8Text generation
Qwen3.8-27B-UD-Q4_K_M.ggufhf:unsloth/Qwen3.8-27B-GGUF/Qwen3.8-27B-UD-Q4_K_M.ggufText generation