LLM Inference Benchmark Explorer
Gemma-4-31B-it on Thor — NVFP4, vLLM, TP1, speculative decoding k=3 inference benchmark
Open this configuration in the LLM Inference Benchmark Explorer
| Model | Gemma-4-31B-it |
|---|---|
| Parameters | 31B |
| Intelligence Index | 14.7 |
| Agentic Index | 4.2 |
| Device | Thor |
| Quantization | NVFP4 |
| Max C | 0 |
| Chat Capacity | 0 |
| Agentic Capacity | 0 |
| TP | — |
| DP | — |
| PP | — |
| Engine | vLLM |
| Speculative Decoding | Yes |
| C | TTFT (ms) | TPS (tok/s) | Status |
|---|---|---|---|
| 1 | 336 | 16.16 | FAIL |
| 2 | 478 | 15.38 | FAIL |
| 4 | 491 | 14.59 | FAIL |
| 8 | 507 | 14.72 | FAIL |
In OpenZeka's measurement, Gemma-4-31B-it (31B parameters), served in NVFP4 format with vLLM and speculative decoding on Thor, reached a generation speed of 16.2 tok/s per request and a time to first token (TTFT) of 336 ms with a single request.
At the default speed targets (at least 20 tok/s per request, TTFT at most 1,000 ms) there is no capacity estimate for this configuration: no measured load level met them.
Notes: Speculative MTP k=3. Served as gemma-4-31b-it. Model: nvidia/gemma-4-31B-it-NVFP4. 256K context
Intelligence Index and Agentic Index values are published by Artificial Analysis and are reproduced here with attribution.
In OpenZeka's measurement, Gemma-4-31B-it (31B parameters), served in NVFP4 format with vLLM and speculative decoding on Thor, reached a generation speed of 16.2 tok/s per request and a time to first token (TTFT) of 336 ms with a single request.
At the default speed targets (at least 20 tok/s per request, TTFT at most 1,000 ms) there is no capacity estimate for this configuration: no measured load level met them.