Can an RTX 5090 run Qwen3.8-27B?
Yes: an RTX 5090 runs Qwen3.8 27B at 75.2 tok/s at Q4_K_M, measured on llama.cpp with 4K context in October 2026.
4 signed runs on llm-speed. Single-request decode speed in tokens per second; the short-chat workload unless noted. All runs were measured by llm-speed (@llmspeed). Runs measured by llm-speed are credited to @llmspeed.
Measured tok/s by quant and runtime
| Quant | Runtime | Build | Decode tok/s (best run) | Workload | Runs | Source |
|---|---|---|---|---|---|---|
| Q4_K_M | llama.cpp | commit 620d2225c | 75.2 | chat-short | 1 | Measured by llm-speed@llmspeed |
| Q4_K_M | llama.cpp | 1 (9725a31) | 65.7median of 3 runs: 65.7 | chat-short | 3 | Measured by llm-speed@llmspeed |
Fastest quant that fits: Q4_K_M on llama.cpp, 75.2 tok/s (measured by llm-speed; it fits because it ran on this hardware). Quants without a run are not listed.
VRAM used: 18.62 GiB runtime-reported GPU allocation for model, context and compute buffers with a 16,384-token context configured (r_c3ps9wjygi9; a runtime-log allocation, not peak device memory).
Every measured row
| Run | Date (UTC) | Workload | Quant | Runtime / build | Prompt tokens | Output tokens | Decode tok/s | Prefill tok/s | TTFT ms | VRAM GiB | Source |
|---|---|---|---|---|---|---|---|---|---|---|---|
| r_qfrb2zmyk0k | 2026-10-15 | chat-short | Q4_K_M | llama.cpp · commit 620d2225c | 154 | 256 | 75.2 | 563.4 | 273 | – | Measured by llm-speed@llmspeed |
| r_c3ps9wjygi9 | 2026-09-05 | chat-short | Q4_K_M | llama.cpp · 1 (9725a31) | 114 | 256 | 65.7 | 973.6 | 117 | 18.62 | Measured by llm-speed@llmspeed |
| r_dz5a3lzxhrj | 2026-09-05 | chat-short | Q4_K_M | llama.cpp · 1 (9725a31) | 114 | 256 | 65.7 | 969.8 | 118 | 18.62 | Measured by llm-speed@llmspeed |
| r_7mod38qsldj | 2026-09-05 | chat-short | Q4_K_M | llama.cpp · 1 (9725a31) | 114 | 256 | 65.4 | 865.3 | 132 | 18.62 | Measured by llm-speed@llmspeed |
| r_qfrb2zmyk0k | 2026-10-15 | chat-long | Q4_K_M | llama.cpp · commit 620d2225c | 3224 | 872 | 74.6 | 2499.5 | 1290 | – | Measured by llm-speed@llmspeed |
| r_7mod38qsldj | 2026-09-05 | chat-long | Q4_K_M | llama.cpp · 1 (9725a31) | 3184 | 874 | 65.1 | 3501.1 | 909 | 18.62 | Measured by llm-speed@llmspeed |
| r_dz5a3lzxhrj | 2026-09-05 | chat-long | Q4_K_M | llama.cpp · 1 (9725a31) | 3184 | 874 | 65.1 | 3496.8 | 911 | 18.62 | Measured by llm-speed@llmspeed |
| r_c3ps9wjygi9 | 2026-09-05 | chat-long | Q4_K_M | llama.cpp · 1 (9725a31) | 3184 | 874 | 65.1 | 3506.5 | 908 | 18.62 | Measured by llm-speed@llmspeed |
| r_qfrb2zmyk0k | 2026-10-15 | agent-trace | Q4_K_M | llama.cpp · commit 620d2225c | 3496 | 840 | 73.5 | 5065.3 | 436 | – | Measured by llm-speed@llmspeed |
| r_qfrb2zmyk0k | 2026-10-15 | concurrent-decode | Q4_K_M | llama.cpp · commit 620d2225c | 1024 | 256 | 75.4 | – | – | – | Measured by llm-speed@llmspeed |
Questions
Can an RTX 5090 run Qwen3.8-27B?
Yes: an RTX 5090 runs Qwen3.8 27B at 75.2 tok/s at Q4_K_M, measured on llama.cpp with 4K context in October 2026.
How many tokens per second does Qwen3.8-27B get on an RTX 5090?
Decode speed measured by llm-speed, by quantization and runtime: Q4_K_M on llama.cpp: 75.2 tok/s; Q4_K_M on llama.cpp: 65.7 tok/s. Based on 4 runs measured by llm-speed; single-request chat-short unless noted.
What is the fastest quant of Qwen3.8-27B that fits on an RTX 5090?
Q4_K_M on llama.cpp was the fastest measured setting at 75.2 tok/s; it fit because it ran. Other quants are listed only where a run measured them.
How much VRAM does Qwen3.8-27B use on an RTX 5090?
The runtime reported 18.62 GiB of GPU allocation for model, context and compute buffers with a 16,384-token context configured. That is a runtime-log allocation, not peak device memory.