Hardware › Strix Halo local LLM benchmarks: 178 measured configs
Strix Halo local LLM benchmarks: 178 measured configs
178 measured local-LLM configs on AMD Strix Halo (Ryzen AI Max+ 395); fastest 14B+ model Ornith-1.5-35B-A3B at 105.6 tok/s. Official specs, Singapore prices read daily, every number linked to its source.
Snapshot 2026-10-16 · 394 configs on AI boxes and 423 on GPU cards · 123 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.
Silicon: AMD Ryzen AI Max+ 395 (Strix Halo). Official specs.
Quick answers
- What is the fastest model on Strix Halo?
- Ornith-1.5-35B-A3B at 105.6 tok/s single-stream (FP4, speculative decode, mtp), source last checked 2026-09-10. Models under 14B parameters are left out of this answer. source
- What is the fastest 100B+ model on Strix Halo?
- Qwen3.8-Flash-Next at 55.3 tok/s single-stream (llama.cpp/hip (speculative decode, mtp)), source last checked 2026-09-18. source
- What is the largest model measured on Strix Halo?
- DeepSeek V4 Flash 284B (284B parameters) at 32 tok/s single-stream (IQ2_XXS, speculative decode, mtp), source last checked 2026-09-16. source
- How many configs does TokenMark track for Strix Halo?
- 178 single-stream configs across 59 models, each linked to the README or model card it came from.
Official specs
- cpu: 16 Zen 5 cores / 32 threads, 3.0 GHz base, up to 5.1 GHz boost, 64 MB L3 (source)
- gpu: AMD Radeon 8060S, 40 RDNA 3.5 compute units (source)
- npu tops: 50 (source)
- max unified memory gb: 128 (source)
- memory type: LPDDR5X-8000, 256-bit bus (source)
- memory bandwidth gbps: 256 (source)
- tdp w: 45-120 W configurable (cTDP); 55 W default (source)
- process node: TSMC 4 nm FinFET (source)
- launch date: Announced CES, January 2025; retail availability early 2025 (source)
Where to buy AMD Strix Halo (Ryzen AI Max+ 395) in Singapore
Read 17 Oct 2026 (SGT) from each seller's own product page. Prices in SGD. Overseas stores are converted at USD/SGD 1.2792 with 9% GST at import added; shipping is not included. Singapore sellers already include GST. List prices on the day, not a quote. No affiliate links.
AMD Ryzen AI Max+ 395 mini PC, 128 GB
- Bosgame (M5) (Official store): S$4,321, ships from overseas
- GMKtec (EVO-X2) (Official store): S$4,880 to S$5,089, ships from overseas
- ALLSTARS (Singapore) (Local distributor, 3yr warranty): Out of stock
- Framework (Desktop) (Official store, configure to order): Price on seller site, ships from overseas
Measured single-stream configs on AMD Strix Halo (Ryzen AI Max+ 395)
| Model | Quant | Backend | Decode | Context | Trust | Source |
|---|---|---|---|---|---|---|
| Qwen3 0.6B | Q8_0 | llama.cpp/vulkan | 266 tok/s (prefill 13112) | 512 | Structured table | source |
| Qwen3 0.6B | Q8_0 | llama.cpp/rocm | 208.7 tok/s (prefill 4666.1) | 512 | Structured table | source |
| LFM2.5-2.6B | Q4_K_M | llama.cpp/vulkan (speculative decode, ngram) | 182.1 tok/s | 8192 | Extracted, source-linked | source |
| LFM2.5 8B-A1B | Q4_K_M | llama.cpp/vulkan | 176.5 tok/s (prefill 3398.4) | 512 | Structured table | source |
| Ornith-1.5-35B-A3B | FP4 | speculative decode, mtp | 105.6 tok/s | Owner-submitted | source | |
| Qwen3-30B-A3B-Instruct-2507 | IQ4_XS | llama.cpp/vulkan | 103.2 tok/s (prefill 1438.1) | 512 | Structured table | source |
| Qwen3-Coder 30B-A3B | Q4_K_S | llama.cpp/vulkan | 98 tok/s (prefill 1406.5) | 512 | Structured table | source |
| Qwen3-Coder 30B-A3B | UD-Q4_K_XL | llama.cpp/vulkan | 97.1 tok/s (prefill 1400) | 512 | Structured table | source |
| NVIDIA Nemotron 3.5 Lightning 30B | FP4 | speculative decode, mtp | 95.2 tok/s | Owner-submitted | source | |
| Qwen3-Coder 30B-A3B | IQ4_XS | llama.cpp/vulkan | 90.4 tok/s (prefill 1372.3) | 512 | Structured table | source |
| Qwen3 30B-A3B NEO-MAX | IQ4_XS | llama.cpp/vulkan | 87.4 tok/s (prefill 1396.1) | 512 | Structured table | source |
| Qwen3.6-35B-A3B | Q4_0 | llama.cpp/vulkan | 81.3 tok/s (prefill 1243.5) | 512 | Structured table | source |
| llm-jp-4.1-32b-a3b-thinking | Q4_K_M | llama.cpp/vulkan (depth-0) | 81 tok/s (prefill 1980.2) | Verified (ran it) | source | |
| Qwen3-30B-A3B | UD-Q4_K_XL | llama.cpp/rocm | 79.1 tok/s (prefill 687.9) | Structured table | source | |
| Nemotron Cascade 2 30B-A3B | IQ4_XS | llama.cpp/vulkan | 79 tok/s (prefill 1325.3) | 512 | Structured table | source |
| Gemma-4-26B-A4B | UD-Q4_K_XL | llama.cpp | 78 tok/s | Owner-submitted | source | |
| Nex N2.5 Mini 35B-A3B | FP4 | 76.9 tok/s | Owner-submitted | source | ||
| Ornith-1.5-35B-A3B | FP4 | 76.9 tok/s | Owner-submitted | source | ||
| Nemotron-3-Nano-30B-A3B | IQ4_XS | llama.cpp/vulkan | 76 tok/s (prefill 1312.5) | 512 | Structured table | source |
| Ornith-1.5-35B-A3B | Q4_K_M | llama.cpp/rocm (speculative decode, mtp-ngram) | 75.9 tok/s (prefill 1306) | Extracted, source-linked | source | |
| Qwen3.5-35B-A3B | IQ4_XS | llama.cpp/vulkan | 75.2 tok/s (prefill 1170.3) | 512 | Structured table | source |
| Gemma 4 26B-A4B IT QAT | UD-Q4_K_XL | llama.cpp/vulkan | 74.8 tok/s (prefill 1432) | 512 | Structured table | source |
| Qwen3.6-35B-A3B | UD-Q4_K_XL | llama.cpp/vulkan (speculative decode, mtp-3) | 74.6 tok/s | Structured table | source | |
| Qwen3-Coder 30B-A3B | UD-Q4_K_XL | llama.cpp/rocm | 73.7 tok/s (prefill 1285.3) | 512 | Structured table | source |
| llm-jp-4.1-32b-a3b-thinking | Q4_K_M | llama.cpp/rocm (depth-0) | 73.6 tok/s (prefill 1416.6) | Verified (ran it) | source |
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.