Created bymario-alka.dePowered bynoob2claw.detricoma.de
Google

Gemma-4-26B-A4B26B

Benchmark profile and published results.

MoE26B
Position in the field

Best values compared

Best metrics of this model against the minimum, average and maximum of all published systems.

Generation264,1 tok/s
Min 0,0Ø 245,7Max 3.484,3
Ueber dem Durchschnitt · 4971 Systeme im Feld
Prefill4.537 tok/s
Min 4Ø 4.892Max 112.346
Unter dem Durchschnitt · 4953 Systeme im Feld
Time to First Token451 ms
Min 20Ø 31.420Max 535.235
Ueber dem Durchschnitt · 4968 Systeme im Feld
Performance profile

Throughput by hardware & engine

Each bubble is a GPU, CPU or engine – position shows prompt processing (X) and output speed (Y), bubble size the number of runs.

GPUby graphics card

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)NVIDIA RTX A6000 - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe)NVIDIA RTX A6000NVIDIA DGX SPARK (GB10) - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)NVIDIA DGX SPARK (GB1...
NVIDIA RTX A6000 264,1 tok/sNVIDIA DGX SPARK (GB10) 47,1 tok/s

CPUby processor

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)AMD Ryzen Threadripper PRO 7955WX 16-Cores - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe)AMD Ryzen Threadrippe...NVIDIA Grace - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)NVIDIA Grace
AMD Ryzen Threadripper PRO 7955WX 16-Cores 264,1 tok/sNVIDIA Grace 47,1 tok/s

ENGby engine

33425016783,50,04.0504.2504.4504.651Prefill (tok/s)Generation (tok/s)llama.cpp - 264,1 tok/s Generation, 4.323 tok/s Prefill, TTFT 2.385 ms (6 Laufe)llama.cppvLLM - 47,1 tok/s Generation, 4.378 tok/s Prefill, TTFT 451 ms (1 Lauf)vLLM
llama.cpp 264,1 tok/svLLM 47,1 tok/s
Throughput & latency

Performance benchmark

Metric:
#Model / MakerMetricsParallelGPU / CPU / RAMRuntime
1Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
264,08 tok/s TG
Prefill 4.424 · TTFT 4.475 ms
10×NVIDIA RTX A6000AMD Ryzen Threadripper PRO 7955WX 16-Cores · 184 GB RAMllama.cppopenclaw_cliQ4_K_M
2Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
223,28 tok/s TG
Prefill 4.432 · TTFT 4.464 ms
10×NVIDIA RTX A6000AMD Ryzen Threadripper PRO 7955WX 16-Cores · 184 GB RAMllama.cppopenclaw_cliQ8_0
3Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
201,09 tok/s TG
Prefill 4.516 · TTFT 2.194 ms
5×NVIDIA RTX A6000AMD Ryzen Threadripper PRO 7955WX 16-Cores · 184 GB RAMllama.cppopenclaw_cliQ4_K_M
4Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
169,09 tok/s TG
Prefill 4.537 · TTFT 2.184 ms
5×NVIDIA RTX A6000AMD Ryzen Threadripper PRO 7955WX 16-Cores · 184 GB RAMllama.cppopenclaw_cliQ8_0
5Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
95,58 tok/s TG
Prefill 4.000 · TTFT 499 ms
1×NVIDIA RTX A6000AMD Ryzen Threadripper PRO 7955WX 16-Cores · 184 GB RAMllama.cppopenclaw_cliQ4_K_M
6Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
87,33 tok/s TG
Prefill 4.028 · TTFT 494 ms
1×NVIDIA RTX A6000AMD Ryzen Threadripper PRO 7955WX 16-Cores · 184 GB RAMllama.cppopenclaw_cliQ8_0
7Gemma-4-26B-A4B26BGoogle Performance benchmarkTimebench 3 - Kombi (Prefill + Generation)
47,06 tok/s TG
Prefill 4.378 · TTFT 451 ms
1×NVIDIA DGX SPARK (GB10)NVIDIA Grace · 120 GB RAMvLLMgodclawNVFP4
Agent & chat rating

Harness benchmark

Noch keine Harnessbenchmarks fuer dieses Modell.