Skip to content

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

C++

Merge f-split-fb4 (201d674e9) into f-split-int: cloudflared discovery, settings file, Gyro minimum engine, two models on two GPUs

Niko1221261145baaf67
diff --cc tools/strata_mcp.py
index 5fb540c79,bcd8fa825..14aac05aa
--- a/tools/strata_mcp.py
+++ b/tools/strata_mcp.py
@@@ -79,12 -79,14 +79,14 @@@ FALLBACK_MODELS =
"vision": True},
"Gyro-S": {"about": "the smallest option, for PCs with little memory (27.6 GB of VRAM + RAM together); lower quality "
"than the 2-3-bit sizes, and the 2-bit Q2_0 decodes faster",
+ "engine": (0, 1, 45, 1),
"download_gb": 58.5, "ram_gb": 27.6, "arena_gb": 25.8, "families": ("gyro",), "mem_total": True,
- "vision": False},
+ "vision": True},
"Gyro-M": {"about": "about IQ3_XXS quality with about 20% less GPU memory (35 GB of VRAM + RAM together); decodes "
"about 20% slower than Q2_0 on 12 GB cards",
+ "engine": (0, 1, 45, 1),
"download_gb": 92.0, "ram_gb": 35.0, "arena_gb": 32.1, "families": ("gyro",), "mem_total": True,
- "vision": False},
+ "vision": True},
"NVFP4": {"about": "4-bit NVFP4 experts, GPTQ (sergqwer's strata-nvfp4 pack); RTX 50 has FP4 tensor cores, other "
"NVIDIA cards run it at 8-bit precision", "download_gb": 128.0, "ram_gb": 80, "arena_gb": 68.0,
"families": ("nvfp4",), "vision": True},