r9700 · gfx1201 · 2 cards
Qwen 3.8 Flash
Tested on our rig UD-Q6_K_XL llama.cpp 262K
llama.cpp Q6, 8 threads. Measured on our rig, 2026-10 — ~16 tok/s writing at 262K.
Launch
llama.cpp is the consumer default. Strata is the Flash try. Exact extra flags stay not recorded when marked.
llama.cpp · Vulkan flags not recorded
# Shape only — llama.cpp Q6, 8 threads on the same Flash file. Exact flags TODO.
./build-vk/bin/llama-server \
-m ~/models/Qwen3.8-Flash-Next-UD-Q6_K_XL.gguf \
-t 8 \
-c 262144 \
--host 127.0.0.1 --port 8080
Gates
Our checklist, in our words — not Sybil’s six-gate wording. Recipe-gate definitions live on methodology. Measured tok/s: How we test.
How we test — measured cells cite run cards; estimates stay labeled.
-
Loads on the named GPU Pass
The process attaches to the discrete card (rocminfo / --list-devices agrees). We do not publish a time-limit SLA.
-
A reply finishes Pass
Generate ends. The server does not hang mid-token. This is not a quality score.
-
Thinking channel (when the model has one) Not measured
A distinct thought block and a final answer both appear. Not measured on our seeded recipes yet.
-
A tool call is used Not measured
The server emits a tool call and consumes the result. Not measured on our seeded recipes yet.
-
Stated context actually ran Pass
The run used the context we printed (262144 for the daily driver). We do not publish a planted-token recall score.
-
Decode recorded Pass
tok/s at the stated context, sourced. No invented floor.
Host RAM
Kit sizes are estimates unless a Local AI RSS number lands.
Minimum system RAM needed
192 GB kit
Estimate · 159.9 GiB raw · 256K context
RAM needed at full context
192 GB kit
Estimate · 159.9 GiB raw · 256K full context
Estimate. Offloaded weights + KV in RAM + runtime + OS/desktop + headroom, rounded up to a 192 GB DIMM kit. The local desktop ships 256 GB DDR5; measured host-RAM usage is not recorded until Local AI numbers arrive.
Breakdown
- Offloaded weights ·
- 125.4 GiB
- Separable embed in RAM ·
- 0 GiB
- KV in RAM at selected ·
- 7.61 GiB
- KV in RAM at full context ·
- 7.61 GiB
- Runtime / backend ·
- 2.88 GiB
- OS / desktop baseline ·
- 8.00 GiB
- Headroom ·
- 16.0 GiB
- Weights
- unsloth/Qwen3.8-Flash-Next-GGUF / UD-Q6_K_XL (folder)
- ROCm / override
- not recorded · none
- Proof
- local desktop · 2026-10 · local desktop strata-test report / Local AI measurements
- Decode / prefill
- 16 / —
Calculator All recipes r9700 page Model page Stack Lab
If the app is not installed, packages are not published yet. There is no public download URL. Read the Stack Lab install notes . If Stack Lab is already running on this machine, open the local app .
AMDai Lab Notes digest
Email only — no account. Double opt-in. About once a week, and only when run cards, guides, ROCm problems, or Stack Lab actually changed. Not a daily drip. Issues may include one labeled Sponsored mention. Slots and rates: Sponsor.