Local recipes

Name: OpenWeights Terminal measured pairs
Creator: Dvidia
License: https://owterminal.com/cite

Every install file. Stacks is the short ranked list. A number is a quote. This desk did not remeasure it.

@MiaAI_lab mia-ai.net/models StacksOne-stream decode, as printed. A dash means the page did not publish one.

Model	Box	How	Quoted	Context	File
GLM 5.3 Flash	×2–4Spark	vLLM · EXL3 4-bit	36.1	1M	GLM-5.3-Flash-EXL3-2x-DGX-Sparks
DeepSeek V4 Flash	×2–3Spark	vLLM · DSpark	38.6	1M	DeepSeek-v4-Flash-DSpark-2x-DGX-Spark
Qwen3.8 27B	×116–32 GB	ExLlamaV3 · EXL3	—	262k	Qwen3.8-27B-16gb-NVIDIA-GPUs-one-click-install
Qwen3.8 Flash Next	×1Spark	vLLM · NVFP4	48.7	262k	Qwen3.8-Flash-Next-Single-DGX-Spark
Qwen3.8 27B	×1Spark	SGLang · NVFP4	18.3	262k	Qwen3.8-27B-SGLang-DGX-Spark
Qwen3.8 Flash Next	×2Spark	vLLM · NVFP4	54.4	262k	Qwen3.8-Flash-Next-Dual-DGX-Sparks
DeepSeek V4.1 Flash	×2Spark	vLLM · EXL3 2.9-bit	31.6	600k	DeepSeek-v4.1-Flash-EXL3-2x-DGX-Sparks
DeepSeek V4.1 Flash	×3–4Spark	SGLang · MXFP4	51.0	256k	DeepSeek-v4.1-Flash-DGX-Sparks
Qwen3.8 27B	×15090	vLLM · NVFP4	~160	262k	Qwen3.8-27B-NVFP4-RTX-5090
Laguna S 2.1	×1Sparkor×1PRO 6000	vLLM · NVFP4	76.7	256k	Laguna-S-2.1-DGX-Spark-RTX-6000-PRO
Qwen3.6 35B-A3B	×1Spark	vLLM · NVFP4	95.1	262k	Unsloth-Qwen3.6-35b-NVFP4-DGX-Spark
Qwen3.8 27B	×1PRO 6000	SGLang · NVFP4	240+	256k	Qwen3.8-27B-RTX-6000-PRO-SGLang-DSpark
MiMo V2.6 Flash	×2Spark	SGLang · MXFP4	35.0	1M	MiMo-V2.6-Flash-2x-DGX-Sparks
Nemotron 3.5 Lightning	×1Sparkor×15090or×1PRO 6000	SGLang · NVFP4	—	1M	Nemotron3.5-Lightning-DGX-Spark-RTX-5090-6000-PRO
Ling 3.0 Flash	×1Spark	SGLang	—	—	Ling-3.0-Flash-SGLang-DSpark-DGX-Spark
Muse Glimmer 30B	×1Sparkor×15090or×1PRO 6000	—	—	—	Muse-Glimmer-30B-DGX-Spark-RTX-5090-6000-PRO