Repository navigation
Basalt on Windows, prebuilt
RTX 50-series only (sm_120): 5090, 5080, 5070, 5060 Ti 16 GB, RTX PRO Blackwell. Nothing older runs it.
Get this ready
- Install CUDA 13 —
basalt.exeneedscublas64_13.dll(any 13.x; built against 13.4). - Install the latest Microsoft Visual C++ v14 redistributable x64 (the 2015-2026 one, 14.51.x, from
visualstudio.microsoft.com/downloads). These were built with MSVC 14.51: with an older redist the exe can fail to
start with a missing-entry-point error onMSVCP140.dll. pip install -r requirements.txt(Python 3.10+).- Download one
.basaltpack from https://huggingface.co/jesdga/Qwen3.8-Flash-Next-Basalt intomodels\. - Edit
basalt_config.json:modelis your pack,gpuis0(or0,1with a second card),portis the web UI.
Run it
bin\basalt.exe serve --config basalt_config.jsonThen open http://127.0.0.1:8080. API: /v1 (OpenAI) and /v1/messages (Anthropic).
System RAM holds the experts that don't fit in VRAM, so leave plenty free. If it starts and then crawls, you ran out.
Other things in here
bin\basalt-probe.exe --model models\your-pack.basalt --engine bin\basalt.exe --fixtures data\bench
bin\basalt-bench.exe --model models\your-pack.basalt --engine bin\basalt.exe --fixtures data\benchbin\basalt.exe serve --help prints everything; docs/running.md and
docs/configuration/index.md explain the knobs.
To build your own: from a Developer command prompt, cmake -S . -B build -G Ninja -DCMAKE_BUILD_TYPE=Release then
cmake --build build.
Checksums
basalt.exe 66f5dad0000457f83db7245c6e3d3cc5d7b0a0b89165c57bc0e6b2ee7efae52e
basalt-bench.exe 73cfdacaea5e2cfc818c270516067eb74ebd3e78a4adc045d12b620aa4cd2b89
basalt-probe.exe ff6a5bffcba5cacf0770f7c2779dd7d3c3f37548496ecdc6df104f2202a2625c
basalt-vision.exe 1728fa5aed2820109c579d404f018b8abeaf4a282f0a400674b0657a949a837d
MIT licence (LICENSE, THIRD_PARTY_LICENSES). Model packs are under the Qwen Community License 1.0.
Windows quality has never been measured: no PPL/KLD or benchmark numbers for it yet.