Qwen3.6-35B-A3B FP8 — radiance container
Qwen/Qwen3.6-35B-A3B-FP8 as a single .rad
container for the radiance inference engine (AMD RDNA4, ROCm), with its MTP head for
speculative decoding and its vision tower.
| File | qwen3.6-35b-a3b-fp8.rad — 34.63 GiB |
| Weights | the published FP8 checkpoint as it ships: every linear, the 256 routed experts included, E4M3 with a scale per 128×128 block; the lm_head made block FP8 by the recipe |
| Speculator | the model's MTP head, which drafts with a 2-bit copy of the lm_head |
| Vision | the 27-block vision tower, bf16: images and video in chat requests |
| Context | 262,144 tokens trained; 200K tested |
Serve
The whole model fits two 32 GB cards.
radiance --model qwen3.6-35b-a3b-fp8.rad --tp 2 --max-model-len 200000 --kv-cache-dtype fp8 \
--max-num-seqs 8 --host 0.0.0.0 --port 8000
The server speaks the OpenAI API (/v1/chat/completions, /v1/completions), with tool calls and
structured output, and image_url / video_url parts in chat
messages. The MTP depth is chosen automatically (--num-speculative-tokens N states one, 0 turns
speculation off).
Built and tested on 2× Radeon AI PRO R9700 (gfx1201).
How this file was made
rad-convert Qwen/Qwen3.6-35B-A3B-FP8 --recipe q36-35b-a3b-fp8.recipe -o qwen3.6-35b-a3b-fp8.rad
The recipe (q36-35b-a3b-fp8.recipe in this repository); everything it does not name is the
checkpoint's own:
output.weight rtn codes=fp8_e4m3 block=128x128 scale=bf16
mtp.draft_head.weight rtn codes=u2 zero=u8 group=128 scale=f16
License
Apache 2.0, as the base model; see LICENSE.