LINKED LIST [txt mode] ▸ Run Qwen 3.8 Flash Next (125B) on con...
home explore | log in

Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s

github.com · first added by @hn_wayback · 2026-10-07 · 2 upvotes

log in to save, upvote or flag this.


─── In 0 lists ─────────────────────────────────────────

(not in any lists yet)

* Qwen 3.8 27B on my laptop can interpret model weights, too (from the discussion)
* DeepSeek V4.1 Flash 1017 (max) - Intelligence, Performance & Price Analysis (from the discussion)
* prabanta-dev/janas: 37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than your RAM, pure C, CPU or GPU. (from the discussion)

─── Discussions ────────────────────────────────────────

* Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s
935 pts · 421 comments · node
see all 2
* Strata: Run a 125B-parameter AI model on a normal gaming PC
3 pts · 1 comment · node

─── From the discussion ────────────────────────────────

* ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
huggingface.co · node
* nixos-config/hosts/zen3-nixos/ai/strata.nix at main · cjdell/nixos-config
github.com · node
* ds4/docs/MODELS.md at main · antirez/ds4
github.com · node
* nathansutton/Qwen3.8-27B-Ternary-Bonsai-2-DFlash2-MLX
huggingface.co · node
* Qwen/Qwen3.8-Flash-Next
huggingface.co · node