Model Releases
If you are using Qwen3.8-27B + MTP, make sure to upgrade to DFlash for extra speed: llama serve -hf ggml-org/Qwen3.8-27B-GGUF --spec-type dr…
If you are using Qwen3.8-27B + MTP, make sure to upgrade to DFlash for extra speed: llama serve -hf ggml-org/Qwen3.8-27B-GGUF --spec-type draft-dflash --spec-draft-n-max 7 Requires the latest llama.cp
If you are using Qwen3.8-27B + MTP, make sure to upgrade to DFlash for extra speed: llama serve -hf ggml-org/Qwen3.8-27B-GGUF --spec-type draft-dflash --spec-draft-n-max 7 Requires the latest llama.cpp v0.6.0 simple: llama serve -hf ggml-org/Qwen3.8-27B-GGUF --spec-type draft-mtp
Related
- on DGX Spark: llama serve -hf ggml-org/Qwen3.8-27B-GGUF:Q4_K_M -hfd ggml-org/Qwen3.8-27B-GGUF:Q4_0 --spec-default --spec-type draft-…
- Pushed: DFlash implementation for llama-cpp. buun-llama-cpp/llama-server -m Qwen3.6-27B.gguf -md dflash-draft-q4_k_m.gguf --spec-type dflash
- llama.cpp recently added DFlash support to its speculative decoding arsenal. Along with MTP, Eagle3 and various ngram-based techniques, the …
- DFlash 2 available for Qwen 3.8 27B and Muse Glimmer
Source: Georgi Gerganov (X) | 2026-10-06