localairecipes launch configs for local inference

Working launch recipes for local LLM inference

Engine configs (vLLM, SGLang, llama.cpp, …) tested on real hardware, with the exact flags, model, quant and context that were used. Copy one, run it, or hand it to your agent. Search or refine below; every recipe on the site is listed here.

Choose your hardware:Pick the cards you own — only recipes that run on them are shown.
✓ CMP 170HX 64GB (Ampere) (32)RTX PRO 6000 Blackwell 96GB (18)DGX Spark 128GB (Blackwell) (14)RTX 3090 24GB (Ampere) (7)Radeon AI Pro R9700 32GB (AMD) (2)RTX 5090 32GB (Blackwell) (2)DGX Station (GB300) 288GB (Blackwell) (1)✓ RTX 4090 24GB (Ada Lovelace) (1)CMP 170HX 64GB (Ampere): Any6 cards (1)4 cards (10)3 cards (3)2 cards (15)1 card (3)RTX 4090 24GB (Ada Lovelace): Any1 card (1)
Refine

Source:

Engine
Model
Quant
Storage requirement
Feature tags

Reset

Filters: 170hx-x3 rtx-4090 clear all

4 recipes · page 1 of 1

Ampere (1)

Community builds

Ada Lovelace (1)

Community build — submitted by users and agents; decode speeds are reported by their authors; verify on your own hardware.

Qwen3.8-Flash-Next IQ3_S + MTP on 1x RTX 4090 — llama.cpp + tandem.cpp MoE expert cache (ewhacc)

by ewhacc - recipe, patch set and measurements↗, llama.cpp / ggml (MIT)↗, MoE expert-split rule adapted from Strata (Niko1221)↗ · listed by localai-recipes

MoE expert-cache offload for llama.cpp via the tandem.cpp patch set: hot experts stay in a GPU expert cache, missed experts are split between CPU compute and an upload into the cac

Qwen3.8-Flash-Next llama.cpp NVIDIA RTX 4090 24GB ×1 (Ada Lovelace) 48 tok/s decode

Ampere (2)

Community build — submitted by users and agents; decode speeds are reported by their authors; verify on your own hardware.

Qwen3.8-Flash-Next Uncensored AWQ-g32 on 3x CMP 170HX — PP3, 1M ctx, NVMe PLE, MTP-4 (givisok)

by recipe, tuning notes and measurements by givisok↗, base stack and vllm-sm80 fork by clee↗, AWQ-g32 checkpoint by leoncca↗, uncensored base by orcarouter↗, cudagraph capture-size idea from mannydelrio↗, vLLM (Apache-2.0)↗ · listed by localai-recipes

MTP-4 production recipe for Qwen3.8-Flash-Next-Uncensored (leoncca AWQ-g32) on 3x CMP 170HX (SM80 cards unlocked to 64 GiB): PP3 partition 17,17,14 (no GPU P2P on this pairing; the

Qwen3.8-Flash-Next PP vllm +2
Qwen3.8-Flash-Next W4A16 3x CMP 170HX PP3 NVMe PLE fp8 KV 1M (clee)

by clee (recipe, fork patches, bench + needle harness)↗, clee (vllm-sm80 fork: units-fix, fp8-kv-sm80, ple-pp3, ple-nvme-staged)↗, Kai / 2Wild (Qwen38-Flash-Next-4x3090 — staged-PLE architecture inspiration)↗, Intel (INT4 W4A16 AutoRound quant)↗, froggeric (bugfixed Qwen3.8 chat templates)↗ · listed by localai-recipes

Working PP3 recipe for Qwen3.8-Flash-Next (Qwen4Exp 180B-A6B MoE, Intel INT4 W4A16 AutoRound) on 3x CMP 170HX (SM80, unlocked 64 GiB each): the 95.4 GiB BF16 PLE n-gram table strea

Qwen3.8-Flash-Next vllm NVIDIA CMP 170HX 64GB ×3 (Ampere) 59.8 tok/s decode · 4690 tok/s prefill