Qwen3.6 reports

generation:

81.1 tokens/s

prompt processing (prefill):

1423.5 tokens/s

quant:

UD-Q4_K_M (GGUF)

kv:

F16

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 81.1 t/s generation and 1423.5 t/s prompt processing on a Mac Studio M4 Max with 128 GB unified memory. Setup is llama.cpp b10794 with UD-Q4_K_M GGUF, flash attention, F16 K/V, single sequence, 99 GPU layers requested. Figures are depth-0 llama-bench microbenchmarks: PP512 and TG128 separately, not end-to-end. At prefix depth 4096 generation falls to 77.79 t/s and at 16384 to 73.18 t/s.

generation:

54.2 tokens/s

prompt processing (prefill):

1049.7 tokens/s

quant:

Q8_0 (GGUF)

kv:

F16

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B-MTP Q8_0 at 54.2 t/s generation and 1049.7 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is llama.cpp build 10603 with Vulkan RADV/Mesa 26.1.7, -ngl 999 -b 2048 -ub 512 -fa on, plain decoding with no speculation, one Radeon 8060S GPU. The headline pairs same-backend Vulkan pp512 1049.70 ±7.40 t/s with tg128 54.22 ±0.09 t/s; the reported SD 0.00 is rounded, not proof of zero variability.

generation:

12.1 tokens/s

prompt processing (prefill):

334.0 tokens/s

quant:

Q4_K_M (GGUF)

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-27B-MTP Q4_K_M at 12.1 t/s generation and 334 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is llama.cpp with Q4_K_M and F16 KV cache, one ROCm/Vulkan device, flash attention on and mmap off, without speculative decoding. The figures are native llama-bench PP512/TG128 means over 3 repetitions at depth 0; the raw logs give 333.95±7.08 t/s and 12.05±0.03 t/s.

generation:

21.5 tokens/s

prompt processing (prefill):

484.0 tokens/s

quant:

BF16 (GGUF)

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 21.5 t/s generation and 484 t/s prompt processing on an AMD Strix Halo 128GB. Setup is llama.cpp with BF16 weights and ROCm flash attention, one GPU device with positive layer offload. The 21.5 t/s and 484 t/s come from separate fixed-work tests, PP512 and TG128, not one chat request; a separate shape 2 run at depth 0 gives 23.71 t/s generation.

generation:

60.4 tokens/s

prompt processing (prefill):

942.2 tokens/s

quant:

UD-Q4_K_M (GGUF)

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B UD-Q4_K_M at 60.4 t/s generation and 942.2 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is llama.cpp with Vulkan, flash attention enabled, one GPU device, at depth 0. The headline run is PP512 942.18±4.92 and TG128 60.39±0.22; PP and TG are separate fixed-work synthetic tests, not one chat request, and the reported values are means over 3 repetitions.

generation:

13.7 tokens/s

prompt processing (prefill):

295.0 tokens/s

quant:

ROCmFP4 STRIX hybrid embF16/headQ6 (GGUF)

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6 at 13.7 t/s generation and 295 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is the charlie12345/rocmfp4-llama fork with ROCmFP4 STRIX hybrid quantization, F16 embeddings and a Q6_K output head, run through native llama-bench with -ngl 999 -fa 1 -ub 1024 -r 3 and no MTP. The 13.7 t/s is the Vulkan0 TG128 figure and 295 t/s the Vulkan0 PP2048 figure, separate synthetic tests rather than one request; a ROCm0 arm measured 13.5 t/s and 335 t/s. Native batch size, KV types, context capacity and resident memory are unpublished.

generation:

13.8 tokens/s

prompt processing (prefill):

294.0 tokens/s

quant:

ROCmFP4 STRIX hybrid embF16/headQ6 (GGUF)

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6 at 13.8 t/s generation and 294 t/s prompt processing on an AMD Strix Halo 128GB rig. Setup is the charlie12345/rocmfp4-llama fork with the ROCmFP4 STRIX hybrid embF16/headQ6 quant, run through native llama-bench with -ngl 999 -fa 1 -ub 1024 -r 3 and no MTP. The 13.8 t/s is the Vulkan0 TG128 figure and 294 t/s the Vulkan0 PP2048 figure, separate synthetic tests rather than one request; batch size, KV types, context depth and resident memory are unpublished.

generation:

71.1 tokens/s

prompt processing (prefill):

1195.0 tokens/s

quant:

ROCmFP4 STRIX hybrid embF16/headQ6 (GGUF)

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6 at 71.1 t/s generation and 1195 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is the charlie12345/rocmfp4-llama fork with the ROCmFP4 STRIX hybrid embF16/headQ6 quant, run through Vulkan0 with -ngl 999 -fa 1 -ub 1024 -r 3 and no MTP. The 71.1 t/s is TG128 and the 1195 t/s is PP2048, separate synthetic tests rather than one request; the same campaign's ROCm0 arm reached 63.4 t/s TG128 and 1271 t/s PP2048.

generation:

12.9 tokens/s

prompt processing (prefill):

126.4 tokens/s

quant:

Q4_K_M (GGUF)

kv:

F16

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6 27B at 12.9 t/s generation and 126.4 t/s prompt processing on an M4 Pro with 64 GB unified memory. Setup is dasLLAMA version 16 with Q4_K_M weights, f16 K/V cache, and one Metal GPU; the run used 512 synthetic prompt positions and 128 serial decode positions. The user notes these are native processing counts, not one 512-input/128-output chat request, and that no speculative decoding was requested.

generation:

66.4 tokens/s

prompt processing (prefill):

787.1 tokens/s

quant:

UD-Q4_K_M (GGUF)

kv:

F16

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 66.4 t/s generation and 787.1 t/s prompt processing on an Apple M4 Pro with 64 GB unified memory. Setup is dasLLAMA version 16 with UD-Q4_K_M weights, f16 K/V cache, and one Metal GPU at 640 allocated positions. The figures are native PP512 and TG128 counts, not a single 512-input/128-output request.

generation:

12.6 tokens/s

prompt processing (prefill):

123.9 tokens/s

quant:

Q4_K_M (GGUF)

kv:

F16

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-27B at 12.6 t/s generation and 123.9 t/s prompt processing on an Apple M4 Pro with 64 GB unified memory. Setup is llama.cpp commit 6fdd0ac89 with the Q4_K_M GGUF, Metal backend, -ngl 99, 10 CPU threads, and f16 K/V cache. The figures are native llama-bench means over five repetitions: PP512 at 123.908607 t/s and TG128 at 12.621998 t/s, each starting from cleared state at depth 0, with load and warmup excluded.

generation:

48.6 tokens/s

prompt processing (prefill):

747.1 tokens/s

quant:

UD-Q4_K_M (GGUF)

kv:

F16

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 48.6 t/s generation and 747.1 t/s prompt processing on an M4 Pro 64GB. Setup is llama.cpp with UD-Q4_K_M, K/V f16, Flash Attention auto, -ngl 99 and 10 CPU threads on one Metal GPU. The figures are native llama-bench PP512 and TG128 means over five repetitions at depth 0, not a single chat request.

generation:

62.0 tokens/s

prompt processing (prefill):

1507.1 tokens/s

quant:

UD-Q4_K_XL (GGUF)

kv:

F16

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 62 t/s generation and 1507.1 t/s prompt processing on a Framework Desktop with AMD Ryzen AI Max+ 395 and Radeon 8060S. Setup is llama.cpp with UD-Q4_K_XL, f16 K/V cache, flash attention, 99 GPU layers on one Vulkan device, no speculative decoding. The figures are separate native tests: PP512 at depth 0 and TG32 at depth 0, not one 512-input/32-output chat. At depth 65536 the same arm measured 617.952691 t/s PP512 and 43.203830 t/s TG32.

prompt processing (prefill):

148.8 tokens/s

quant:

UD-Q4_K_XL (GGUF)

kv:

F16

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-27B UD-Q4_K_XL prefill at 148.8 t/s on an AMD Strix Halo 128GB machine. Setup is llama.cpp via lemonade ROCm 7 gfx1151 with f16 K and V cache, flash attention on, batch 2048 and 16 CPU workers, single Radeon 8060S with full layer offload. The figure is the 131072-token input cell of a PP-only sweep with two measured repetitions per cell and depth 0.

prompt processing (prefill):

32.1 tokens/s

quant:

UD-Q4_K_XL (GGUF)

kv:

F16

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-27B UD-Q4_K_XL at 32.1 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is llama.cpp with f16 K and V cache, flash attention on, batch 2048 and 16 CPU workers, one Radeon 8060S with full layer offload. The figure is the 131072-token prefill cell of a PP-only sweep with depth 0 and n_gen 0; the source's exact mean is 32.145328 t/s, and no decode rate is supplied.

generation:

64.0 tokens/s

prompt processing (prefill):

1005.3 tokens/s

quant:

Q4_K_M (GGUF)

kv:

F16

flash attention:

on

mtp (multi-token prediction):

off

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 64 t/s generation and 1005.3 t/s prompt processing on an AMD Strix Halo 128GB machine. Setup is llama.cpp with Q4_K_M, f16 K and V cache, flash attention on, batch 2048, microbatch 512, and 16 CPU workers. The figures are the quiet-mode baseline of a five-cell campaign, measured as separate PP512 and TG128 tests, not a 512-in/128-out chat.

generation:

61.0 tokens/s

prompt processing (prefill):

1323.0 tokens/s

quant:

UD-Q4_K_XL (GGUF)

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is input throughput, not time to first token. Check the full setup before comparing.

User reports Qwen3.6 35B-A3B at 61 t/s generation and 1323 t/s prompt processing on one Strix Halo integrated GPU. Setup is llama.cpp Vulkan pin 7fe450e with Unsloth Dynamic UD-Q4_K_XL GGUF, a 20.8 GiB file; KV cache datatype and layer offload are unreported. The figures are separate synthetic llama-bench phases, pp512 and tg128, one run per configuration, not client TTFT or end-to-end throughput; total context capacity is unspecified.

generation:

124.5 tokens/s

quant:

4-bit (MLX)

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is shown only when the source reports it; generation speed cannot tell us time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 124.5 t/s on a single Mac Studio with an M5 Max and 36 GB of unified memory. Setup is LM Studio's MLX backend with MLX 4-bit weights; configured context capacity, KV precision and Flash Attention are unknown. The figure is the article's rounded single-stream decode rate, measured as client-observed streaming decode-phase throughput rather than engine-native timing, and a separate four-stream run reached 265.6 t/s aggregate.

generation:

145.8 tokens/s

quant:

4-bit (MLX)

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is shown only when the source reports it; generation speed cannot tell us time to first token. Check the full setup before comparing.

User reports Qwen3.6 35B-A3B at 145.8 t/s on a single Mac Studio with an M5 Max and 36 GB of unified memory. Setup is oMLX 0.7.0 with MLX 4-bit weights and an SSD cache directory enabled; KV precision, context capacity and speculative decoding are unknown. The 145.8 t/s is a rounded single-stream decode rate from a developer comparison, not engine-native timing; a separate four-stream run reached 281.9 t/s aggregate.

generation:

173.4 tokens/s

quant:

4-bit (MLX)

mtp (multi-token prediction):

on

Reported by the source; GPU count, offloading and concurrent requests can change these figures. Prompt-processing speed is shown only when the source reports it; generation speed cannot tell us time to first token. Check the full setup before comparing.

User reports Qwen3.6-35B-A3B at 173.4 t/s on a single Mac Studio with an M5 Max and 36 GB of unified memory. Setup is Rapid-MLX 0.16.1 default with MLX 4-bit weights and MTP speculative decoding on the text-only lane; configured context capacity, KV precision and Flash Attention are unknown. The figure is the article's rounded single-stream decode rate, client-observed streaming decode-phase throughput rather than engine-native timing, and the source warns differences under about 5% may be run variation.