Language

Qwen 3.8 27B DFlash-2 Drafter

DFlash-2 block-diffusion speculative drafter for Qwen3.8-27B. Drafts a whole block per pass and keeps the top candidates at every position. Pair via `--spec-type draft-dflash`.

Apache 2.0Source verified

Specification

Model ID

qwen3.8-27b-dflash

Family

qwen3.8

Modality

Language

Parameters

DFlash-2 drafter

Context

262,144 tokens

Quantization

Q4_K_M

Weights

1.1 GB

Minimum RAM

4 GB

Source

Weights and provenance.

Repository

z-lab/Qwen3.8-27B-DFlash2-GGUF

Registry license

Apache 2.0

Source license

apache-2.0

Access

Open — weights fetch without accepting additional terms.

Serve it

# Pull the weights onto a node
tenzro model download qwen3.8-27b-dflash

# Serve it
tenzro model serve qwen3.8-27b-dflash

# Or query the registry entry over JSON-RPC
curl https://rpc.tenzro.xyz \
  -X POST -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,
       "method":"tenzro_getModel",
       "params":[{"model_id":"qwen3.8-27b-dflash"}]}'

Same family

Qwen 3.8 2.4T-A95B (MoE)

2.4T (MoE, 95B active)