Configs › Qwen3.5-35B-A3B › Qwen3.5-35B-A3B on Strix Halo: 75.2 tok/s, 3 measured configs
Qwen3.5-35B-A3B on Strix Halo: 75.2 tok/s, 3 measured configs
Measured Qwen3.5-35B-A3B speeds on Strix Halo: fastest 75.2 tok/s single-stream (IQ4_XS, llama.cpp/vulkan), 3 configs with quant, backend and context, each linked to its source, snapshot 2026-10-29.
Snapshot 2026-10-29 · 413 configs on AI boxes and 288 on GPU cards · 121 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.
Fastest measured single-stream run of Qwen3.5-35B-A3B on Strix Halo: 75.2 tok/s (IQ4_XS, llama.cpp/vulkan), snapshot 2026-10-29.
Every measured single-stream run of Qwen3.5-35B-A3B on Strix Halo
| Decode tok/s | Quant | Backend | Context | Trust | Measured by | Source checked | Source |
|---|---|---|---|---|---|---|---|
| 75.2 | IQ4_XS | llama.cpp/vulkan | 512 | Structured table | hogeheer | 2026-10-14 | source |
| 64.9 | Q4_K_M | llama.cpp/vulkan | 512 | Structured table | hogeheer | 2026-10-14 | source |
| 54.7 | Q4_K_M | llama.cpp/rocm | 512 | Structured table | hogeheer | 2026-10-14 | source |
Best quant on Strix Halo
Fastest plain-decoding run per quant, versus the fastest quant.
- IQ4_XS: 75.2 tok/s, fastest (llama.cpp/vulkan)
- Q4_K_M: 64.9 tok/s, 14% lower (llama.cpp/vulkan)
Backends on Strix Halo
Fastest plain-decoding run per backend. Quants can differ between rows.
- llama.cpp/vulkan: 75.2 tok/s (IQ4_XS)
- llama.cpp/rocm: 54.7 tok/s (Q4_K_M)
Qwen3.5-35B-A3B on other boxes
- DGX Spark: 131 tok/s best of 1 config
Questions and answers
- How fast is Qwen3.5-35B-A3B on Strix Halo?
- The fastest of 3 measured configs is 75.2 tok/s single-stream (IQ4_XS, llama.cpp/vulkan). The slowest is 54.7 tok/s.
- Which quant of Qwen3.5-35B-A3B is fastest on Strix Halo?
- IQ4_XS at 75.2 tok/s, ahead of Q4_K_M at 64.9. Plain decoding only.
- Which backend is fastest for Qwen3.5-35B-A3B on Strix Halo?
- llama.cpp/vulkan at 75.2 tok/s, then llama.cpp/rocm at 54.7. Each backend figure is its best plain-decoding row and may use a different quant.
- How does Strix Halo compare with other boxes for Qwen3.5-35B-A3B?
- Strix Halo best 75.2 tok/s. Other boxes, best measured: DGX Spark 131. Backends and quants differ, so this is the best known number per box, not a controlled test.
See also: Qwen3.5-35B-A3B on every box · Strix Halo hardware page (from S$4,329 in Singapore today).
Last updated: snapshot 2026-10-29; newest source check 2026-10-14.
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.