# Throughput Benchmark — glm53

- **Run ID**: `2026-09-22T07-56-30.349354Z_feffd52f`
- **Date**: `2026-09-22T07:56:30.349678+00:00`
- **Mode**: throughput-only
- **tool-eval-bench**: `v2.7.0`

## Run Context

| Parameter | Value |
|---|---|
| Backend | vllm |
| Server | `http://***:30000/v1` |
| Model (API) | `glm53` |
| Model (Root) | `local-inference-lab/GLM-5.3-Flash-NVFP4` |
| Temperature | 0.0 |
| Seed | 42 |
| Max Turns | 8 |
| Timeout | 120.0s |
| Scenarios | all (88) |
| Parallel | 1 (sequential) |
| Error Rate | 0.0 |
| Thinking | enabled |
| Max Tokens | 16384 |

## Inference Engine

| Property | Value |
|---|---|
| Engine | vLLM 0.26.1rc0+glm53.r39.vllm9e5d1793 |
| Max Model Length | 524,288 |
| Host | `localhost` |
| Platform | `macOS-26.6.2-arm64-arm-64bit-Mach-O` |
| Python | 3.14.7 |

## Results

| Test | pp t/s | tg t/s | TTFT (ms) | Total (ms) | Tokens |
|---|---:|---:|---:|---:|---:|
| pp2048 tg128 @ d0 | 2,018 | 58.6 | 951 | 3,106 | 2048+128 |
| pp2048 tg128 @ d0 c2 | 2,149 | 69.2 | 1,286 | 4,414 | 2048+128 |
| pp2048 tg128 @ d4096 | 2,498 | 59.4 | 2,231 | 4,358 | 2048+128 |
| pp2048 tg128 @ d4096 c2 | 2,468 | 67.4 | 4,049 | 7,249 | 2048+128 |
| pp2048 tg128 @ d8192 | 2,515 | 58.7 | 3,600 | 5,754 | 2048+128 |
| pp2048 tg128 @ d8192 c2 | 2,489 | 51.4 | 6,593 | 10,253 | 2048+128 |