qwen38

Optimized SGLang runtime for Qwen3.8-27B FP8 with DFlash2 and Qwen3.8 Flash-Next NVFP4 with FR-Spec on one NVIDIA RTX PRO 6000 Blackwell 96 GB GPU (SM120): 524K context, HiCache and NIXL.

Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K prompt), full 256K window at 2,865 tok/s; with 64 GB RAM 3,410 tok/s prefill · 84 tok/s decode. New: opt-in uncensored mode (runtime abliteration, no new weights).

A 198t/s and 353t/s qwen 3.8 27b inferencing engine for 2*v100 and 4*v100

Qwen3.8-Flash-Next 177B MoE (IQ3_S 3.44bpw) on RTX 5090 Laptop 24GB: full-model quality @ 97.7 tok/s · dual-lane 256K full-context unlocked on Strata 0.1.50 · 486K tokens in flight · 60+ measured rounds, 5 quant tiers · 中英双语实测实录

Run Qwen3.8-Flash-Next on ONE RTX 3090 (24 GB) + 64 GB RAM: 128K context, up to 2,100 tok/s prefill, 43–51 tok/s decode. vLLM runtime with hot MoE experts on the GPU and cold experts computed on the CPU, INT8 KV cache, Docker, OpenAI-compatible API.

Swift 1.5 post-train of Qwen3.8-Flash-Next on 4x RTX 3090 with vLLM: W4A16 checkpoint on the unchanged Flash-Next v2.5.1 engine, TP2 x PP2 + EP, 924,993-token FP8 KV pool, 262K context, MTP K=3, recalibrated KV scales.

Qwen3.8-27B and the ThinkingCap fine-tune at 262K context on dual or quad RTX 3090s: Merlin engine v2.5.2 + 11 patches, FP8 KV pool 826,689 (dual) / 1,974,072 (quad), MTP K=4

海光 K100-AI（gfx928）上跑通并优化 Qwen3.8-Flash-Next INT8 W8A8：4 卡 TP4，单流 31 tok/s、20 并发聚合 175 tok/s，含 262K 上下文、MTP 投机解码、工具调用与完整实测

Inference engines for Gemma, Qwen and Pocket TTS in pure Go. No cgo, no Python. One static binary, CPU or Vulkan.

Browser-native paged inference for Qwen3.8 27B and Gemma 3. Quantized WebGPU kernels, bounded weight streaming, UI, and browser JavaScript SDK.

What is Improve Qwen3

Qwen 3.8 builds The Backrooms

Lean-QwenFlash: Qwen3

MTP (Model-Translation-Postprocess) is a ground-breaking technology designed to enhance the capabilities of Qwen, a sophisticated language model

Qwen3

Local deployment of Ternary Bonsai 2 27B (PrismML 1.75-bit ternary quantized Qwen3.8 27B) on an RTX 4060 laptop — setup, benchmarks and notes

A tower defense made in Three js to test Qwen3.8-27B performance ( hf download hf://unsloth/Qwen3.8-27B-GGUF/Qwen3.8-27B-UD-Q3_K_XL.gguf , 128K context )

Run Qwen3.8-27B locally in Claude Code Desktop with vision, tools, native reasoning and Claude Code CLI support. Validated on an NVIDIA RTX 5090 32 GB.

Qwen3