A llama.cpp expert-streaming build runs the 177B Qwen 3.8 Flash Next MoE at roughly 11.5 tok/s on a 12GB RTX 5070 and 32GB DDR4-2400 by streaming experts from SSD.
GPU model
Capacity range
Total VRAM
Practice
Model
Quantization
Context length
Reported speed
Runtime
Tools / workflow
Record origin
EXPLORE BY CAPACITY
Start with your machine.
Capacity is the reported total for this setup. Multi-GPU and unified-memory layouts can behave differently.
8 GB or lessLaptops, iGPUs and modest cards9–12 GBEntry-level local inference13–16 GBCurrent mid-range cards17–20 GBRX 7900 XT and compact pro cards21–24 GB3090, 4090 and 7900 XTX class25–32 GB5090 and workstation cards33–48 GBWorkstation and multi-GPU setups49–96 GBLarge multi-GPU and datacenter setups97 GB+Unified memory, servers and clusters
3 resultsEvery record keeps its origin visible.
CodingGenesis archive2026-10
Qwen3.8 Flash Next 177B at 11-15 tok/s on a single RTX 5070 12GB with 32GB DDR4-2400Research & analysisGenesis archive2026-09
Qwen 3.8 Flash Next on 12GB of VRAM via RAM+VRAM offload (Strata engine)Four quants of a vision model measured on a 12GB RTX 5070 SFF with 64GB DDR5, through a purpose-built inference engine, at 37 to 55 GB combined RAM+VRAM.
CodingGenesis archive2026-09
The first local model this person actually trusts with real workOn a 12GB 5070, a heavily quantized Qwen 3.8 27B through a custom Hermes backend is, in their words, the first local model that's felt confidently productive.