Qwen3.6-27B on RTX 5090 (32GB) + AMD EPYC 7543 32-Core Processor (120c) + 945GB
MeasuredOn 2026-10-17, Qwen3.6-27B Q4_K_M ran at 75.1 tok/s (chat-short) on an RTX 5090 (32GB) with llama.cpp. Run r_pob-co51bd-.
RTX 5090 (32GB) + AMD EPYC 7543 32-Core Processor (120c) + 945GB
suite suite-v1
cli 0.0.7
signedWh0Zhp7kVD…
Embed badgesubmitted Oct 17, 2026Workload results
| Workload | Backend | Model | decode tok/s | prefill tok/s | TTFT | p50 | p95 |
|---|---|---|---|---|---|---|---|
| chat-short | llama.cpp@0.7.0-dev (build 1, commit 1268eda) | Qwen3.6-27BQ4_K_M | 75.09tok/s | 302.2tok/s | 371ms | 13.2ms | 13.6ms |
| chat-long | llama.cpp@0.7.0-dev (build 1, commit 1268eda) | Qwen3.6-27BQ4_K_M | 74.55tok/s | 2,644.8tok/s | 1,203ms | 13.4ms | 13.7ms |
| concurrent-decode | llama.cpp@0.7.0-dev (build 1, commit 1268eda) | Qwen3.6-27BQ4_K_M | 75.13tok/s | — | — | 13.3ms | 13.7ms |
| agent-trace | llama.cpp@0.7.0-dev (build 1, commit 1268eda) | Qwen3.6-27BQ4_K_M | 74.13tok/s | 4,464.5tok/s | 485ms | 13.3ms | 14.1ms |
Reproduce on your machine
Start with this model and workload selection. This command does not capture the original artifact, backend version or server settings:
Match quantization, context capacity, caching and runtime settings for a useful comparison. Duration varies by workload and hardware; the published client may differ from this run's version. How it's measured.
Share this run
Copy the short link, a Reddit-ready summary, or a README badge. Every number is the measured result on this page, and each badge links back to it.
https://llm-speed.com/r/r_pob-co51bd-Related benchmarks
- More Qwen3.6-27B benchmarks — every backend and rig that has run this model.
- More RTX 5090 (32GB) LLM benchmarks — every model measured on this hardware.
Provenance
- Run ID
- r_pob-co51bd-
- Fingerprint hash
- Public key
- Wh0Zhp7kVDuyBDd8x/XiH27gtQqNlro/TMJEqvvbcvo=
- Received
- 2026-10-17 09:55:33