Summary
Qwen3.8-Flash-Next is an experimental open-weight multimodal model previewing the architecture planned for Qwen4: a 125B-parameter LM with 6B activated, plus a 51B n-gram embedding table and a 4B multi-token-prediction head (~180B total in BF16). Attention is hybrid — Gated DeltaNet linear attention with Qwen Sparse Attention every 4th layer — with 262,144 tokens of native context extensible to 1M via YaRN. The card benchmarks it ahead of Qwen3.8-27B and DeepSeek-V4-Flash-0731 on agentic suites (DeepSWE 1.1 58.7, SWE-bench Pro 62.5, LiveCodeBench v6 91.9, GPQA Diamond 91.7). It ships as post-trained weights for Transformers, vLLM and SGLang under the custom qwen-community-1.0 license, with the official Qwen3.8-Flash positioned as its production counterpart.
Why it matters
For serving teams this is the second frontier-lab open-weight release in a week to bet on hybrid linear attention for long-context cost — GLM-5.3-Flash shipped the same idea on 8/25. Treat it as an architecture signal rather than a production drop: it is explicitly experimental, the license is custom rather than MIT/Apache, and the production Qwen3.8-Flash with 1M default context is the supported counterpart.
Technical details
| Architecture | 125B LM (6B activated) + 51B n-gram embedding (bigram/trigram lookup) + 4B multi-token prediction; ~180B total params BF16; 48 layers in repeating pattern 3x (Gated DeltaNet -> MoE) then 1x (Qwen Sparse Attention -> MoE) |
|---|---|
| Attention | hybrid: Gated DeltaNet linear attention (48 V / 16 QK heads, dim 128) + Qwen Sparse Attention every 4th layer (24 Q / 2 KV heads, dim 256; indexer budget 512 blocks / 2048 tokens) |
| Moe | 512 experts, 10 routed + 1 shared per token; Gated Residual 4 branches, bottleneck rank 320 |
| Context | 262,144 native; extensible to 1M via YaRN |
| Modality | image-text-to-text multimodal; thinking mode default (effort xhigh/medium/low) |
| Training | Muon + AdamW split across weight categories; no batch-size warmup; larger learning rates |
| Benchmarks | DeepSWE 1.1 58.7, SWE-bench Pro 62.5, SWE-bench Multilingual 81.0, NL2Repo-Bench 48.1, Agents' Last Exam 51.2, GPQA Diamond 91.7, LiveCodeBench v6 91.9; vision: AndroidWorld 84.5, LVBench 76.6 — vs Qwen3.8-27B, Qwen3.7-Plus, DeepSeek-V4-Flash-0731, Claude-Opus-4.6 |
| License | custom qwen-community-1.0 (license: other) — check terms before commercial use |
| Adoption | 52,341 downloads (30d) + 4,287 likes @2026-08-30T00:00Z; FP8 variant 44,281 downloads; production counterpart = Qwen3.8-Flash on QwenCloud/Model Studio/OpenRouter since launch week ($0.15/$0.47 per M) |
Updates
2026-08-30 Production counterpart verified (resolves the prior watch item): Qwen3.8-Flash is the API-only production version of Flash-Next — live on QwenCloud since launch week with a 1M-token default context and official built-in tools; served via Alibaba Cloud Model Studio (USD 0.15/M input <=1M tokens, 0.47/M output, context-caching discount) and OpenRouter (0.15/0.47). No separately licensed open weights beyond Flash-Next itself (custom qwen-community-1.0). Adoption re-check 2026-08-30: main repo 52,341 downloads + 4,287 likes (was 4,213 on 8/29; the 30d download pool is unchanged at 52,341 — treat as a static window, not zero growth). Recommendation stays WATCH (experimental preview; production path is the API).
2026-08-31 Adoption check @2026-08-31T00:00Z: 121,976 downloads (+133% vs the 52,341 @8/30 sample, ~7 days after listing) + 4,384 likes (+97) — the display counter finally moved after appearing frozen between the 8/29 and 8/30 samples. Sustained-velocity data point for trend #1 criterion (c). Recommendation stays WATCH.
2026-09-02 adoption check @2026-09-02T00:0xZ: 207,941 downloads (+70% vs 121,976 @8/31, ~9 days after listing) + 4,635 likes; FP8 variant 130,451; unsloth GGUF derivative 431,339 (repo created 8/26). slotstream (Show HN 9/1, 147 pts) runs the 104GB 4-bit weights on a 48GB M5 Pro at ~12 tok/s warm decode (32GB peak, SSD streaming, single Swift binary, Ollama/OpenAI-compatible APIs) — independent community serving tooling, though still not a benchmark reproduction. Architecture now documented in the Qwen3.8-Next design paper (arXiv 2608.30320, ev-20260901-03): 125B/6B active + 51B n-gram embeddings, ~1/9 training FLOPs vs the 397B-A17B predecessor at comparable quality. Stays WATCH
2026-09-05 Adoption check @2026-09-05: 351,374 downloads (+69% vs 9/2) + FP8 repo 186,676 (+43%) + 4,871 likes; unsloth/Qwen3.8-Flash-Next-GGUF at 702,251 — now exceeds Qwen's own main repo; nvidia/Qwen3.8-Flash-Next-NVFP4 created 9/2 (NVIDIA publishing an NVFP4 derivative); mudler APEX GGUF derivative 9/4. Quantization recipe research catching up: Minima NVFP4 W4A4 across a Qwen3.8-27B hybrid incl. GDN layers (ev-20260905-03). Stays WATCH
2026-09-14 2026-09-14 report run: SGLang main merged a reproducible single-DGX-Spark path for NVIDIA's NVFP4 derivative (#39126). The narrow 200-question GSM8K check reported 97.5%/97.0% with/without MTP and 92.87/89.53 tok/s, but this is TP=1, main-branch-only and not a controlled comparison. Recommendation stays WATCH.
Tags
open-weightsarchitecture-previewlinear-attentionmoelong-contextmultimodal