LLM Inference Benchmark Explorer
Gemma-4-26B-A4B-it on 1× DGX Spark — FP16, vLLM, TP1, speculative decoding inference benchmark
Open this configuration in the LLM Inference Benchmark Explorer
| Model | Gemma-4-26B-A4B-it |
|---|---|
| Parameters | 26B |
| Intelligence Index | 16.7 |
| Agentic Index | — |
| Device | 1× DGX Spark |
| Quantization | FP16 |
| Max C | 2 |
| Chat Capacity | 8 |
| Agentic Capacity | 3 |
| TP | — |
| DP | — |
| PP | — |
| Engine | vLLM |
| Speculative Decoding | Yes |
| C | TTFT (ms) | TPS (tok/s) | Status |
|---|---|---|---|
| 1 | 256 | 30.58 | PASS |
| 2 | 668 | 23.46 | PASS |
| 4 | 448 | 19.38 | FAIL |
| 8 | 510 | 15.42 | FAIL |
| 16 | 707 | 12.25 | FAIL |
| 32 | 744 | 10.20 | FAIL |
| 64 | 901 | 8.35 | FAIL |
In OpenZeka's measurement, Gemma-4-26B-A4B-it (26B parameters), served in FP16 format with vLLM and speculative decoding on 1× DGX Spark, reached a generation speed of 30.6 tok/s per request and a time to first token (TTFT) of 256 ms with a single request.
Considering the speed targets and the available KV cache capacity, the estimated capacity is 8 users for chat use and 3 for agentic use. The realistic capacity will likely fall between these two values. In scenarios dominated by coding, tool use, long workflows and multi-agent use, capacity approaches the agentic estimate; where shorter interactions, standard conversations and lighter tasks dominate, it approaches the chat estimate.
Intelligence Index and Agentic Index values are published by Artificial Analysis and are reproduced here with attribution.
In OpenZeka's measurement, Gemma-4-26B-A4B-it (26B parameters), served in FP16 format with vLLM and speculative decoding on 1× DGX Spark, reached a generation speed of 30.6 tok/s per request and a time to first token (TTFT) of 256 ms with a single request.
Considering the speed targets and the available KV cache capacity, the estimated capacity is 8 users for chat use and 3 for agentic use. The realistic capacity will likely fall between these two values. In scenarios dominated by coding, tool use, long workflows and multi-agent use, capacity approaches the agentic estimate; where shorter interactions, standard conversations and lighter tasks dominate, it approaches the chat estimate.