llamaperf

AMD Strix Halo 128GB

AMD · 128GB unified memory · 338 reports

On the AMD Strix Halo 128GB, Qwen3.8 125B · 6B active at 4-bit, the model most run on it, generates a median of 26.6 tokens per second from 8 plain runs (community reports, as of 11 Nov 2026).

As of 11 Nov 2026, the models most run on the AMD Strix Halo 128GB, with generation and prompt-processing medians from plain runs (one device, one request, no speculative decoding):

Engines people use on it: llama.cpp 98 (reports) · Gufo 52 (reports) · Strata 4 (reports) · vLLM 3 (reports) · LM Studio 2 (reports) · ds4 2 (reports)

Run models on your AMD Strix Halo 128GB? Add your numbers.

Your coding agent can run a speed test on your machine and send it here with a free account. Or paste a result you've already posted.

Reported performance on the AMD Strix Halo 128GB

These reports mix GPU counts and memory setups, so a model listed here may have needed extra cards or CPU offloading. The calculator estimates which models fit in 128 GB of unified memory. Prompt processing is how fast the input is read, not how long you wait for the first word. How to measure time to first token.

Showing 20 of 338 reports. See all of them →

Tone: mixedAMD hardware
generation:
34.1 tokens/s

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 Qwen3.8-Flash-Next at 34.1 tok/s decode on a Ryzen AI Max+ 395 128GB mini PC. Setup is Gufo at 16K context, with the model weights taking around 103.7 GiB. With MTP decode rises to 53.7 tok/s, and adding the NPU gives 34.2 tok/s decode with about 10% faster prefill. The user questions whether the largest model is the best use of the memory.

Nov 11, 2026
Tone: positiveAMD hardware
generation:
62-68 tokens/s
prompt processing (prefill):
1900-2100 tokens/s
quant:
FP4

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.

agenticcodingtool-uselong-context

User reports Qwen3.8 Flash-Next 125B FP4 decoding at 62-68 tok/s on a Strix Halo 128GB mini PC (Ryzen AI Max+ PRO 495) under real agent traffic. Setup is the Halogen engine with a ~54GB FP4 checkpoint plus a ~48GB n-gram table and MTP drafter, 262k context and 4 concurrent slots; cold 28k prefill ran at ~1.9-2.1k tok/s and prompt cache hit rate was 98-99.6%. MiniMax-M3.1-Flash (428B, 93GB IQ1_S/IQ2_XXS quant) loaded but decoded at about 1 tok/s because 31% of experts streamed from NVMe; GLM-5.3-Flash ran at ~21 tok/s and Qwen3-30B-A3B at 70-93 tok/s. The model scored 73.3/100 on the user's Godot agent benchmark.

Nov 10, 2026
Tone: positiveAMD hardware
generation:
27.1 tokens/s
prompt processing (prefill):
873.0 tokens/s
quant:
ud-iq3_xxs (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 Qwen3.8-Flash-Next at 27.1 tok/s decode (45.6 tok/s with drafting) on a single AMD Strix Halo 128GB in performance mode with 113GB RAM dedicated to the GPU. Setup is the libosmos engine with the ud-iq3_xxs pack; prompt processing runs 873/822/811/709 tok/s at 2k/8k/16k/64k steps. The engine was tuned at a 70W power envelope. The post also lists DeepSeek V4.1 Flash 1017 decode of 19.5/42.5, 17.5/35.4 and 14.4/26.5 tok/s plain/drafted for ud-iq1_s, ud-iq1_m and ud-iq2_xxs, with 2k/8k prompt speeds of 380/534, 373/519 and 316/429 tok/s.

Nov 9, 2026
AMD hardware
generation:
34.1 tokens/s
prompt processing (prefill):
1888.0 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 compares Gufo against stock llama.cpp on a 128GB Strix Halo (Ryzen AI Max+ 395) running the full UD-Q4_K_XL Qwen3.8-Flash-Next, with FCLK locked at 2000 MHz. Gufo was run with its internal bench and GUFO_PREFILL_TILE=16384; figures are after warm up, three runs. Gufo reached 1888 t/s on a 16K prefill, 35.0 t/s decode at short context and 34.1 t/s decode at 16K context.

Nov 9, 2026

Unknown family

AMD Strix Halo 128GB · Gufo · 16,384 ctx

Tone: mixedAMD hardware
generation:
60.9 tokens/s
prompt processing (prefill):
1888.0 tokens/s

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 benchmarks Gufo on a Strix Halo 128GB mini PC (Nimo 395), splitting work between GPU and NPU with FCLK locked at 2000 MHz. At 16k context, GPU-only prefill reaches 1888 t/s and GPU+NPU 2080 t/s without MTP; with MTP the figures are 1812 t/s and 2002 t/s, roughly +10% either way. Short-context decode stays at 60.9 t/s with or without the NPU. Higher 16k decode numbers in some GPU+NPU+MTP runs are not treated as a pure NPU speedup because acceptance rate and output change. The model is not named.

Nov 9, 2026
Tone: positiveAMD hardware
generation:
41.0 tokens/s
prompt processing (prefill):
1240-1335 tokens/s
quant:
UD-Q4_K_XL (GGUF)
kv:
q8_0

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.

codingagenticlong-context

User reports Qwen3.8-Flash-Next decoding at about 41 tok/s at 55-67K context and 36-48 tok/s at 230K on a Strix Halo 128GB laptop (96 GB GPU / 32 GB Windows split), with cold prefill of about 1,240-1,335 tok/s. Setup is Gufo v0.31 with UD-Q4_K_XL weights, q8_0 KV cache, two sessions of 262,144 context, the shared Q8_0 MTP head with 7 draft tokens, and spill-mode disk cache; about 98.4 GB of GPU memory at load. The patches cut per-turn restore at 221-231K from 3.9-4.9 s to 0.67-1.0 s, and a 237K chat spills 3.69 GB to disk in 15.7 s. MTP acceptance is 70-91%. No formal f16-vs-q8 KL measurement was done, and testing was on one Windows machine only.

Nov 9, 2026
AMD hardware
generation:
70-80 tokens/s

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.

coding

User asks how fast Qwen3.8 Flash-Next runs on a Bosgame M5 128GB Strix Halo, with Gufo or Strata, and which quant to use. No hardware was run and no engine or quant is settled; the user is considering the machine. The only figure is secondhand, 70-80 t/s that others mention, and the user asks whether that holds in everyday coding use.

Nov 7, 2026

Qwen3.8 27B

AMD Strix Halo 128GB · Gufo · 262,144 ctx

Tone: positiveAMD hardware
generation:
24.0 tokens/s
prompt processing (prefill):
450.0 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.

coding

User reports Qwen3.8-27B at 24 t/s generation and 450 t/s prefill at 262,144 context on a Strix Halo box. Setup is recent Gufo builds with UD-Q4-K-XL weights and DFLASH speculative decoding, using around 40GB of memory. User says speed stays steady as context grows, unlike earlier llama.cpp and MTP attempts that fell to single digits. They ask about automatically switching between QFN for planning and coding and the 27B for memory-heavy builds, noting QFN is faster but uses about 102GB of RAM.

Nov 6, 2026
AMD hardware
generation:
38.0 tokens/s
prompt processing (prefill):
480.0 tokens/s
quant:
UD-IQ1_S (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 DeepSeek-V4.1-Flash 1017 at up to 38 tok/s decode and over 480 tok/s prompt processing on a single Strix Halo 128GB (Ryzen AI MAX+ 395, Radeon 8060S / gfx1151). Setup is the libosmos engine with unsloth UD-IQ1_S weights plus an MXFP8 engram, with drafting enabled for the decode figure. The figures are stated as ceilings ("up to"), and a huihui abliteration is said to be available through the engine.

Nov 5, 2026
Tone: positiveAMD hardware
generation:
21.0 tokens/s
prompt processing (prefill):
520.0 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.

codingagenticlong-context

User reports Qwen3.8-Flash-Next at about 21 t/s decode at 4K context and about 14 t/s at 128K on an AMD Strix Halo 128GB, with prefill 520 t/s falling to 230 t/s. Setup is llama.cpp (llama-server) on ROCm with UD-Q4_K_XL, 262K context and the MTP draft head, driven headless by Claude Code. MTP gives about 1.5x up to 64K and becomes a loss past about 100K. The review ran 53 turns over 41 minutes with context growing to about 110K tokens; the user calls it an anecdote, not a benchmark.

Nov 3, 2026
Tone: positiveAMD hardware
generation:
60.0 tokens/s
prompt processing (prefill):
1200.0 tokens/s

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.

agenticlong-context

User reports Qwen3.8 Flash-Next at 60 t/s generation and 1200 t/s prefill on one of four Strix Halo 128GB boxes. Setup details such as engine, quant and KV cache are not stated. The four boxes are used at one, two, three and four boxes for different models. On two boxes MiMo Flash runs at 20 t/s with 420 t/s prefill and GLM Flash at 16 t/s with 400 t/s prefill. DeepSeek 4.1 Flash on three boxes and GLM-5.3 on four boxes are not yet tested. The user hopes to raise MiMo prefill to 800 and GLM Flash to 20+ t/s with 600-700 prefill.

Nov 2, 2026
Tone: positiveAMD hardware
generation:
99.8 tokens/s
prompt processing (prefill):
1648.0 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.

tool-use

User reports Qwen3.8-Flash-Next at 99.82 t/s MTP decode and 1,648 t/s prefill at 32K context on a Strix Halo machine running Windows 11. Setup is Gufo-win v0.5.0 with UD-Q4_K_XL weights and an MTP Q8_0 draft, greedy decoding, two repetitions, one process at a time. Autoregressive decode without MTP was 32.07 t/s at 32K; the table spans 8K to 256K, where MTP decode falls to 90.03 t/s and prefill to 1,508 t/s. Cached follow-up turns return a first token in 21-23 ms at 32K and 27-30 ms at 100K, against 8.6-9.0 ms and 102-113 ms for Bubo 0.1.1 on the same machine.

Nov 2, 2026
Tone: positiveAMD hardware
generation:
40.0 tokens/s
quant:
UD-Q4_K_XL (GGUF)
kv:
q8_0

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.

agenticlong-context

User reports about 40 tok/s single-user with Qwen3.8 Flash-Next UD-Q4_K_XL on a Strix Halo laptop with 128 GB unified memory under Windows. Setup is Gufo v0.26.0 with three patches, q8_0 KV cache and MTP, serving two sessions of 262,144 context to two LAN users; live KV for both sessions uses 98.5 GB of GPU memory versus about 105 GB at f16. Combined throughput with both users generating is about 55-58 tok/s. One chat reached 261,107 tokens with cache=memory on every turn, first token in about 5-6 s at 250K+ instead of a 4-5 minute re-read, and about 250-280K combined context was held in the RAM cache. Full re-reads still occur when an agent is steered mid-task or the client restarts.

Nov 1, 2026
Tone: positiveAMD hardware
generation:
78.0 tokens/s
prompt processing (prefill):
4100.0 tokens/s
quant:
Q8_0 (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.

summarization

User reports LFM3-1.7B at 4.1k t/s prefill and about 78 t/s decode on a Strix Halo iGPU. Setup is llama.cpp with Q8_0 GGUF weights on ROCm. Tensor-splitting across a 7900 XTX plus XT gives about 7.1k t/s prefill and about 151 t/s decode. User finds it more coherent at 32k-64k prompts than a 1.7B model should be.

Nov 1, 2026
Tone: positiveAMD hardware
generation:
25.9 tokens/s
prompt processing (prefill):
1277.0 tokens/s

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 Laguna S 2.1, a ~118B MoE with 256 experts (10 routed plus 1 shared), running end to end on a Strix Halo / Ryzen AI MAX+ 395 with a gufo fork using the NPU alongside the GPU. Decode is 25.9 tok/s plain and prefill reaches up to 1,277 tok/s with the NPU helping, versus ~1,145 tok/s GPU-only. Setup is the gufo engine with NPU+GPU offload; quant and context length are not stated. With the DFlash v1 drafter decode rises to 44-45 tok/s on code and JSON, while prose is more like 28 tok/s, as acceptance depends on content type.

Nov 1, 2026
AMD hardware
generation:
35.0 tokens/s
prompt processing (prefill):
1888.0 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 Qwen3.8-Flash-Next at 35.0 t/s short-context decode on a Strix Halo 128GB (Ryzen AI Max+ 395, Radeon 8060S). Setup is Gufo with UD-Q4_K_XL GGUF, GPU only, no MTP, 128 GB unified memory with about 117 GiB GTT and a 512 MiB BIOS VRAM reservation. At 16K context prefill is 1888 t/s and decode 34.1 t/s. The 103.69 GiB model spans four GGUF shards and fits without issue.

Oct 31, 2026
Tone: positiveAMD hardware
generation:
43.0 tokens/s
prompt processing (prefill):
1539.0 tokens/s
quant:
AWQ W4A16 (AWQ)

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 Qwen3.8-Flash-Next at 43 tok/s decode and 1,539 tok/s prefill on a Strix Halo 128GB at a 5.6k prompt. Setup is a vLLM fork with the libstrix plugin, AWQ W4A16 weights, 16k prefill chunks and MTP 2, single stream. The 102 GB n-gram table is streamed from disk and prefetched into the page cache. At 26.9k and 100.5k prompts decode is 43 and 40 tok/s with prefill 1,617 and 1,572 tok/s. Before this work the same setup did 684-806 tok/s prefill and about 18 tok/s decode. KLD against stock is 0.008-0.019 with 97-98% top-1 agreement, perplexity within 0.6%, and GSM8K 93 vs 92.

Oct 31, 2026
Tone: positiveAMD hardware
generation:
35.0 tokens/s
prompt processing (prefill):
1888.0 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 Qwen3.8-Flash-Next at 35.0 t/s short-context decode and 1888 t/s prefill on a Strix Halo 128GB (Ryzen AI Max+ 395) box. Setup is Gufo with UD-Q4_K_XL GGUF weights (103.69 GiB across four shards), GPU only, no MTP, FCLK locked at 2000 MHz, ROCm 7.2.3. At 16K context decode is 34.1 t/s. Stock llama.cpp ROCm HIP on the same machine gave around 568 t/s prefill and 23-24 t/s decode. The model is the ~125B MoE with ~6B active plus n-gram embedding and MTP modules, fitting in the 128 GB pool with about 117 GiB GTT reported.

Oct 30, 2026
Tone: positiveAMD hardware
generation:
24.5 tokens/s
prompt processing (prefill):
1228.0 tokens/s
quant:
IQ1_M (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.

codingvisionlong-context

User reports DeepSeek V4.1 Flash at 24.5 t/s decode and 1,228 t/s prefill on a single Strix Halo with their Gufo fork. Setup is Gufo with the IQ1_M GGUF resident, HTTP with real chat prompts, greedy decoding and thinking off; DSpark speculation reaches 30.7 t/s on mixed text and 73.6 t/s on repetitive text at depth 0. At 32K depth decode is 24.1 t/s with 1,177 t/s prefill, and at 128K 23.1 t/s with 1,045 t/s prefill. Eight concurrent users get 86.1 t/s total plain and 73.8 t/s with DSpark on mixed text. IQ2_XXS now runs with experts streamed from a Samsung 990 PRO SSD, at 20.8 t/s decode and 1,064 t/s prefill at depth 0.

Oct 30, 2026
Tone: positiveAMD hardware
generation:
54-79 tokens/s
prompt processing (prefill):
1556.0 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 Qwen3.8 Flash-Next at 1,556 tok/s prefill on a Geekom A9 Mega (Ryzen AI Max+ 395, 128 GB LPDDR5X), with decode swinging 54-79 tok/s as draft acceptance varies between 53% and 80%. Setup is Halowin Bubo, a Gufo fork for Windows 11 and Strix Halo, running the UD-Q4_K_XL GGUF at 108.6 GiB across 8 files, with 96 GB carved out to the GPU in BIOS and a fixed 128 GB page file. Context window is 262K. User says it does not slow down over long context; a 20.3M-token session at a 98% cache hit never pushed the live window past 73%. Load takes 24-26 s. The engine is three weeks old, Windows-only by design, and the user is not affiliated with its authors.

Oct 30, 2026

Get a weekly email of new AMD Strix Halo 128GB reports and newly released models that fit it.

Email me new reports

Reported configurations

Individual observations from this page, not expected speeds or a ranking. Select a model to open its report, with the source and offloading conditions. Reported context may be a limit rather than actual input length.

Model / hardwareQuant / engineReported contextGeneration
Qwen3.8 125B (6B active) Flash-Next
AMD Strix Halo 128GB
Not reported
Gufo
16,38434.1 tokens/s
Qwen3.8 125B (6B active) Flash-Next
AMD Strix Halo 128GB
ud-iq3_xxs
libosmos
Not reported27.1 tokens/s
Qwen3.8 125B (6B active) Flash-Next
AMD Strix Halo 128GB
UD-Q4_K_XL
Gufo
16,38434.1 tokens/s
Unknown family
AMD Strix Halo 128GB
Not reported
Gufo
16,38460.9 tokens/s
Qwen3.8 125B (6B active) Flash-Next
AMD Strix Halo 128GB
UD-Q4_K_XL
Gufo
262,14441.0 tokens/s
Qwen3.8 27B
AMD Strix Halo 128GB
UD-Q4-K-XL
Gufo
262,14424.0 tokens/s

Frequently asked

How fast is the AMD Strix Halo 128GB for local LLMs?

On the AMD Strix Halo 128GB, Qwen3.8 125B · 6B active at 4-bit, the model most run on it, generates a median of 26.6 tokens per second from 8 plain runs (community reports, as of 11 Nov 2026). Also reported: Qwen3.6 35B · 3B active at 4-bit (60.0 t/s, 7 plain runs); Qwen3-Coder 30B · 3B active at 4-bit (69.9 t/s, 6 plain runs).

What is the biggest model reported running on the AMD Strix Halo 128GB?

The largest model with a plain generation run on the AMD Strix Halo 128GB is DeepSeek V4.1 Flash 552B · 16B active at 1-bit (24.5 t/s, one plain run), as of 11 Nov 2026. Plain means one device, one request, no speculative decoding, so a run split across several cards does not count.