Qwen3.8-Flash-Next-W4A16-Merlin

<div align="center"> <img src="https://huggingface.co/halt95/Qwen3.8-Flash-Next-W4A16-Merlin/resolve/main/assets/hero.png" alt="Qwen3.8-Flash-Next W4A16, powered by Merlin: a 4-bit checkpoint, 115 GiB, full 262K context on four 24 GB GPUs. A 924,993-token KV pool in VRAM; 638 tok/s total for 8 agents at once; 184 tok/s decode at 262K; 6,051 tok/s prefill on a 100K prompt." width="100%"> <p> <a href="https://huggingface.co/Qwen/Qwen3.8-Flash-Next"><img src="https://img.shields.io/badge/base-Qwen3.8--Flash--Next-615ced" alt="Base model: Qwen3.8-Flash-Next"></a> <a href="#the-checkpoint"><img src="https://img.shields.io/badge/quant-W4A16-3987e5" alt="Quant: W4A16"></a> <a href="#what-it-fits-on"><img src="https://img.shields.io/badge/context-262K-ffb000" alt="Context: 262K"></a> <a href="#what-it-fits-on"><img src="https://img.shields.io/badge/GPUs-Ampere-76B900" alt="GPUs: NVIDIA Ampere (Ada profiles ship untested)"></a> <a href="https://github.com/halt95/merlin-engine"><img src="https://img.shields.io/badge/runs_on-Merlin_3.0.0-a78bfa" alt="Runs on Merlin 3.0.0"></a> <a href="https://huggingface.co/halt95/Qwen3.8-Flash-Next-W4A16-Merlin/blob/main/LICENSE"><img src="https://img.shields.io/badge/licence-Qwen_Community_1.0-lightgrey" alt="Licence: Qwen Community 1.0"></a> </p>

<b><a href="#the-checkpoint">The checkpoint</a> · <a href="#speed">Speed</a> · <a href="#quality">Quality</a> · <a href="#what-it-fits-on">What it fits on</a> · <a href="#use-it">Use it</a> · <a href="#the-base-model">The base model</a> · <a href="https://github.com/halt95/merlin-engine/blob/main/models/flash-next/benchmarks/2026-10-20/BENCH-CARD.md">Bench card</a></b>

</div>

Qwen3.8-Flash-Next-W4A16-Merlin

A 4-bit build of Qwen3.8-Flash-Next, Qwen's 125B mixture-of-experts model (6B active per token, with vision), made to run at its full 262,144-token context on four 24 GB GPUs. It is 115 GiB instead of the original 335 GiB: the routed experts are INT4, the 51B-parameter n-gram table is FP8, and attention and the vision tower stay in BF16. It is served by the Merlin engine, a vLLM 0.32.0 fork tuned for Ampere and Ada that keeps the n-gram table in host RAM and the whole KV cache in VRAM. Measured on RTX 3090s; the engine also ships profiles for other Ampere and Ada cards, from RTX 4090s and 48 GB workstation cards to A100s and CMP 170HX (see What it fits on).

_{numbers: Merlin 3.0.0 on 4× RTX 3090 w/o NVLINK @ 220W, P2P on, 2026-10-20 bench card. KV pool: FP8, all in VRAM; three 262K sessions held at once. Agents: eight 16K conversations, then eight simultaneous 512-token follow-ups, thinking off. Decode: one stream, thinking on (effort low, as benchmarked), exactly 261,888 prompt tokens + 128 out, median of 3. Prefill: one cold 101,875-token prompt. Each figure is also in a table below.}

<a href="https://ko-fi.com/halt95"><img src="https://storage.ko-fi.com/cdn/kofi3.png?v=3" alt="Buy me a coffee on Ko-fi" height="40"></a>

The checkpoint

Files in this repository

_{115.3 GiB in total. There are no shards 00002 and 00016: model.safetensors.index.json maps no tensor to them, so nothing is missing.}

Speed

<img src="https://huggingface.co/halt95/Qwen3.8-Flash-Next-W4A16-Merlin/resolve/main/assets/speed.png" alt="Speed on 4x RTX 3090, Merlin 3.0.0. Decode, one stream: 168 tok/s at 4K, 170 at 131K, 184 at 262K. Cold prefill: 6,081 tok/s at 32K, 6,046 at 131K, 5,838 at 262K. Time to first token: 0.79 s at 4K, 21.6 s at 131K, 44.9 s at 262K. Agents at once, total: 222 tok/s for 2, 378 for 4, 638 for 8. The same figures are in the tables below." width="100%">

_{4× RTX 3090, the needle-prompt ladder (4,076 / 32,596 / 130,416 / 261,888 prompt tokens), every send a cold prefill except at 262K, where sends 1 and 2 are cold and send 3 reuses the prefix before the needle. Decode: thinking on, ≤ 256 tokens out (128 at 262K), median of 3. Prefill: prompt tokens / time to first token (the median below 262K; the cold send 1 at 262K). The 262K needle is found in 3 of 3 thinking-off sends.}

_{One cold request per prompt, 8 tokens out; the 261,888-token row is the exact-262K prompt's cold send 1 (128 tokens out). The 101,875-token figure is the hero's prefill number.}

_{N 16K conversations, then N simultaneous 512-token follow-ups, thinking off, total tok/s while all N decode. Each overlap window was short (4.55, 5.22 and 5.97 s), so read these rows loosely.}

_{N needle prompts (thinking off, greedy) sent at once, then re-sent one by one: every re-send served from the prefix cache means the pool held all N. Peak pool use 100 %, 0 preemptions.}

<details> <summary><b>Against the previous release (v2.5.3)</b></summary>

_{tok/s. Decode: thinking on, median of 3, same frozen driver and prompts. Prefill: one cold send each. v2.5.3 added the Swift1.5-Huihui checkpoint and kept the v2.5.2 engine unchanged, so its Flash-Next figures are the v2.5.2 bench card's. The exact 261,888-token row (184 tok/s) is new in 3.0.0 and has no v2.5.3 counterpart.}

</details> <details> <summary><b>How it was measured</b></summary>

• Host: 4× RTX 3090 (24 GB, sm_86) at 220 W, PCIe Gen4 with P2P, no NVLink, driver 610.43.02.
• Engine: Merlin 3.0.0 from the release wheel, installed with release/install-env.sh, served by merlin serve with the release defaults, one fresh boot.
• Layout: TP2 × PP2 + EP on four cards, MTP K=3, FP8 KV, BF16 draft lm_head, thinking effort low, the default when benchmarked on 2026-10-20 (Merlin now serves medium by default).
• Decode: T=0, a needle prompt at depth, every send a cold prefill except send 3 at 262K, which reuses the prefix before the needle; one discarded warm-up, then 3 sends per cell, median. tok/s = tokens after the first streamed event / time from first to last event.
• Streams are client-timed. Full tables, thinking-off rows and ITL: the bench card.

</details>

Quality

This checkpoint, measured on the Merlin engine:

_{GSM8K and images: the Merlin 3.0.0 release wheel, 2026-10-20. Needle panel: a pre-release engine build, 2026-10-17. These are checks of this checkpoint, not a matched BF16 comparison: no BF16 run of the same checks was made, so how much quality the 4-bit build keeps is not quantified here.}

The base model, as reported by Qwen for the BF16 weights:

_{A selection from the upstream card, which has the full tables and comparisons with other models. These were not re-run on this 4-bit build.}

What it fits on

_{Measured on the reference host (2026-10-20 bench card): three 262K and six 131K sessions held at once with every re-send served from cache; a 4096 × 2048 image (8,388,608 px) with at least 1,026 MiB free on every card.}

Other GPUs. The engine also ships profiles for these; the Status column says how far each has been run:

_{merlin serve and ./setup.sh pick the profile from the GPUs they find. A ~ is the engine's estimate, not a measurement. "(8 at once)": the KV pool holds 17 full 262K conversations, and the server admits 8 requests at a time; the rest queue. On sm_80 cards (A100, A30, CMP 170HX) this checkpoint runs a BF16 KV cache, since its attention path takes FP8 KV only on sm_86 and sm_89 for now, so fewer sessions fit; FP8 there comes in a 3.0.x update. 🟢 measured by us (the 4× CMP 170HX row on a rented host) · ⚫ a profile ships, not yet run by us (expected to work). Every profile: the engine's hardware matrix.}

Use it

git clone https://github.com/halt95/merlin-engine.git && cd merlin-engine
./setup.sh --model flash-next

One command does the rest: after a small uv and Python bootstrap it checks the GPUs, driver, RAM and disk, shows the engine, package and checkpoint downloads with their sizes before fetching them, installs the prebuilt engine, downloads and verifies this checkpoint, and starts an OpenAI-compatible server on 127.0.0.1:8000; requests name the model flash-next-v2 (also flash-next). Enter takes the recommended answer at each question; --yes takes them all, --dry-run shows the plan. Run ./setup.sh again to start the server, and git pull && ./setup.sh to update the engine without downloading the model again.

[!IMPORTANT] This checkpoint needs the Merlin engine; stock vLLM was not tested with it.

<details> <summary><b>Manual install</b></summary>

git clone https://github.com/halt95/merlin-engine.git && cd merlin-engine
curl -LO https://github.com/halt95/merlin-engine/releases/download/merlin-v3.0.0/vllm-0.32.0+merlin.7002f56537-cp313-cp313-linux_x86_64.whl
MERLIN_WHEEL="$PWD/vllm-0.32.0+merlin.7002f56537-cp313-cp313-linux_x86_64.whl" \
  bash release/install-env.sh ./venv

venv/bin/hf download halt95/Qwen3.8-Flash-Next-W4A16-Merlin --revision cec53e9f90e5995b81691604ec0b32191debf797 \
  --local-dir ~/models/flash-next
(cd ~/models/flash-next && sha256sum -c "$OLDPWD/models/flash-next/checkpoint.sha256")
venv/bin/python merlin_cli.py serve ~/models/flash-next

MERLIN_WHEEL installs the prebuilt engine (sm_80, sm_86, sm_89), checked against release/merlin-wheel.sha256; without it the installer compiles the engine instead, about 1.5 hours on 8 cores. merlin serve recognises the checkpoint, picks the largest tested layout for the GPUs it finds (or, with none, the largest untested one that fits, labelled as such) and serves on 127.0.0.1:8000. Add --dry-run to print the resolved command without starting anything. The first start compiles kernels.

</details>

Recommended settings (Qwen's, for this model):

_{Qwen's recommendation for this model, with min_p=0.0 and repetition_penalty=1.0 in both modes. A presence_penalty between 0 and 2 reduces endless repetition; higher values can mix languages.}

• Turn thinking off for a request with "chat_template_kwargs": {"enable_thinking": false}. With thinking on, a short max_tokens can end inside the reasoning with empty content.
• Reasoning effort: Merlin serves medium by default; pick another per request with "chat_template_kwargs": {"reasoning_effort": "low"} (low / medium / xhigh).
• Chat template: Merlin serves Froggeric's fixed Qwen3.8 template (v22.5) by default, not the file in this repository. It accepts system messages after the first turn, so coding agents' per-turn notes stay in place and the prefix cache keeps hitting. CHAT_TEMPLATE=<path> serves another template; CHAT_TEMPLATE= (empty) serves this repository's own.
• With Qwen's own template (CHAT_TEMPLATE=), earlier turns' reasoning is kept by default and "chat_template_kwargs": {"preserve_thinking": false} keeps only the latest.
• Images: up to 42 per request (the launcher's cap), 8 MP each; four 8 MP images in one prompt were checked on the Flash-Next checkpoint.
• Tool calls use the standard tools field, parsed with the Qwen3 coder tool parser.
• To reach it from other machines, answer network at setup's Access question (it generates an API key), or start merlin serve with HOST=0.0.0.0 and VLLM_API_KEY set; put a TLS reverse proxy in front before opening it beyond your own network.

The base model

Qwen3.8-Flash-Next is Qwen's preview of the architecture behind Qwen4. What changed, in short:

• Hybrid attention. Three Gated DeltaNet (linear-attention) layers for every Qwen Sparse Attention layer. QSA attends over micro-blocks instead of single tokens, which keeps long-context latency down.
• Sparse experts. 512 experts per layer with 10 routed and 1 shared active per token: 6B of the 125B parameters do the work for each token.
• N-gram embedding. A 51B-parameter bigram/trigram table at layer 2 adds parameter capacity with less computation than adding experts would. Merlin keeps it in host RAM, so the VRAM goes to KV cache.
• Gated residual. Widened residual streams with a data-dependent read gate and a per-branch write gate.
• Multi-token prediction. One MTP layer, trained over several steps; Merlin uses it for speculative decoding.
• Thinking by default, with a non-thinking mode per request; Qwen's template keeps earlier turns' reasoning by default (preserved thinking).

_{Architecture as reported by Qwen on the upstream card.}

Other builds

_{One stream, thinking on, 261,888 prompt tokens + 128 out, median of 3; Merlin 3.0.0, 2026-10-20 bench cards. Effort as each bench card ran it: Flash-Next low; Swift1.5, Swift1.5-Huihui and ThinkingCap xhigh; Qwen3.8-27B medium. Merlin now serves medium on every Qwen3.8 checkpoint by default. Nemotron and Ornith 1.5 ran at full thinking: neither template reads reasoning_effort. Ornith 1.5 was measured on 2026-10-21 on the reference host with its engine changes (uniform MoE verify on) before they merged into 3.0.0.}

Good to know

• Every figure we measured for this checkpoint comes from one setup: 4× RTX 3090 at 220 W, PCIe Gen4 x16 with P2P, except the rented-host row under Other GPUs; upstream results and the Other builds rows use their stated setups. Hosts without P2P use the no-P2P profiles (see What it fits on).
• Only if you shrink the KV pool yourself: keep it larger than --max-model-len plus a few blocks. With the shipped settings this cannot occur.
• Coming from the qwen38-flash-next-3090s engine (v2.x or earlier)? It is replaced by the Merlin engine; this checkpoint's files are unchanged.

Changes

_{Card-only edits are not listed. Engine releases and their speed changes: the engine's CHANGELOG.}

Credits and licence

Qwen (model; Qwen Community License 1.0, see LICENSE), Intel (AutoRound experts), RadixArk (FP8 PLE table), DominikBucko (MTP packing recipe), froggeric (the fixed Qwen3.8 chat template the engine serves), the vLLM community PLE-offload work, aikitoria (peer-to-peer on consumer Ampere), Neiko2002, and a community idle-power report (behind the v2.5.2 engine fix). Upstream fixes are credited in the engine repository.