{"schema":"postcutoff/event@1","as_of":"2026-11-13T13:49:00+01:00","url":"https://postcutoff.com/e/2026-08-26-qwen3-8-flash-next/","md":"https://postcutoff.com/e/2026-08-26-qwen3-8-flash-next/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-08-26-qwen3-8-flash-next","date":"2026-08-26","date_precision":"day","short_title":"Qwen3.8-Flash-Next, a 125B MoE with 6B active, previews the Qwen 4 architecture","deck":null,"takeaway":null,"category":"open-source","category_label":"Open source","importance":3,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":[{"n":1,"title":"Qwen blog - Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency","url":"https://qwen.ai/blog?id=qwen3.8-flash-next","type":"official","group":"primary","domain":"qwen.ai"},{"n":2,"title":"Qwen3.8-Flash-Next tech report (GitHub)","url":"https://github.com/QwenLM/Qwen3.8-Flash-Next/blob/main/tech_report.pdf","type":"paper","group":"primary","domain":"github.com"},{"n":3,"title":"Hugging Face - Qwen/Qwen3.8-Flash-Next","url":"https://huggingface.co/Qwen/Qwen3.8-Flash-Next","type":"code","group":"primary","domain":"huggingface.co"},{"n":4,"title":"GitHub: Strata, Qwen3.8-Flash-Next on consumer hardware (community)","url":"https://github.com/Niko1221/Strata","type":"code","group":"primary","domain":"github.com"},{"n":5,"title":"Bloomberg: Alibaba releases smaller, cost-effective Qwen AI model","url":"https://www.bloomberg.com/news/articles/2026-08-26/alibaba-releases-smaller-cost-effective-qwen-ai-model","type":"press","group":"press","domain":"bloomberg.com"},{"n":6,"title":"MarkTechPost: Qwen3.8-Flash-Next technical breakdown","url":"https://www.marktechpost.com/2026/08/26/alibabas-qwen-team-releases-qwen3-8-flash-next-a-125b-multimodal-moe-with-6b-active-parameters-previewing-the-qwen4-architecture/","type":"press","group":"press","domain":"marktechpost.com"},{"n":7,"title":"The Decoder: Qwen3.8-Flash-Next targets ultimate cost efficiency","url":"https://the-decoder.com/alibaba-releases-qwen3-8-flash-next-targeting-ultimate-cost-efficiency/","type":"press","group":"press","domain":"the-decoder.com"},{"n":8,"title":"Hacker News: Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s (430 pts)","url":"https://news.ycombinator.com/item?id=49953495","type":"discussion","group":"reactions","domain":"news.ycombinator.com"}],"official":4,"filed":"2026-09-29","updated":"2026-10-04","orgs":["Alibaba","Qwen"],"title":"Qwen3.8-Flash-Next: 125B MoE with only 6B active previews Qwen 4 architecture","summary":"Alibaba's Qwen team open-sourced Qwen3.8-Flash-Next on 2026-08-26: a 125B-parameter multimodal MoE activating just 6B parameters per token, with n-gram embeddings and hybrid Gated DeltaNet/sparse attention, explicitly positioned as a preview of the Qwen 4 architecture; Bloomberg said it rivals Claude Opus 4.6 and DeepSeek V4-Flash.","key_facts":["125B backbone + 51B n-gram embeddings + 4B multi-token-prediction = ~180B on disk; 6B active per token","512 experts, 10 routed + 1 shared per token; Gated DeltaNet in 3 of 4 layers + Qwen Sparse Attention","Context: 262,144 native, 1M with YaRN","Reported benchmarks: SWE-bench Pro 62.5, AndroidWorld 84.5, MathVision 95.7","Training cost ~1/9 of Qwen3.7-Plus; up to 7.6x prefill and 4.9x decode speedup at 1M tokens","License: qwen-community-1.0 (not Apache 2.0)","Local inference: the community project Strata (github.com/Niko1221/Strata, MIT, created Sept 24, 2026; ~10.5k stars by Oct 4) packages 2–3-bit GGUF quants with a one-click installer for 12 GB+ NVIDIA/AMD cards and 32–64 GB RAM; its README reports 94 tok/s output on an RTX 5070 (Q2_0); it reached 430 points on Hacker News on Oct 4","Strata's claims are self-reported: an HN commenter measured much worse vision accuracy than llama.cpp with the same weights, and others questioned quality below 4-bit; the submitter reported 124 tok/s on an RTX 4090 with 128 GB RAM"],"key_numbers":[],"tags":["llm","open-weights","china","moe","efficiency"],"science":null,"body_md":"## What happened\n\nWeights for Qwen3.8-Flash-Next landed on Hugging Face and ModelScope (BF16 and FP8) on 2026-08-26. The model combines an extreme sparsity ratio (6B of 125B active),\na 20M-entry n-gram embedding table, and linear-attention (Gated DeltaNet) layers interleaved with sparse attention — the Qwen team presented it as an early look at Qwen 4 so developers can prepare tooling.\n\n## Why it matters\n\nIt pushes the cost frontier: near-frontier agentic coding numbers at 6B active parameters make strong models cheap to serve at 1M-token contexts.","disputed":[],"related":[{"id":"2026-09-21-dettmers-dlab-open-source-week","url":"https://postcutoff.com/e/2026-09-21-dettmers-dlab-open-source-week/","date":"2026-09-21","date_precision":"day","short_title":"Tim Dettmers' dlab open-source week: 1.5-bit inference and a 125B model on one 24 GB GPU","deck":null,"takeaway":"If the claims hold, 1.5-bit weights and single-consumer-GPU inference of 100B+ MoE models push open-weights AI further out of datacenters.","category":"open-source","category_label":"Open source","importance":2,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":5,"official":3,"filed":"2026-09-30","updated":"2026-09-30","orgs":["dlab","Carnegie Mellon University"]},{"id":"2026-08-03-alibaba-qwen3-8-max","url":"https://postcutoff.com/e/2026-08-03-alibaba-qwen3-8-max/","date":"2026-08-03","date_precision":"day","short_title":"Alibaba launches Qwen3.8-Max and open-sources the Qwen3.8 family","deck":null,"takeaway":"Together with Kimi K3 and DeepSeek V4, Qwen3.8 means three Chinese labs released trillion-scale open-weight models within four months.","category":"model-release","category_label":"Model releases","importance":4,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":5,"official":1,"filed":"2026-09-29","updated":"2026-09-29","orgs":["Alibaba","Qwen"]}],"people":[],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-10-04","type":"edit","text":"Added the community Strata local-inference project (HN front page, 430 pts) and caveats from the HN thread"},{"date":"2026-09-30","type":"edit","text":"Added the official Qwen blog post, tech report and Hugging Face links"},{"date":"2026-09-29","type":"filed","text":"Created"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-09-29","run":null,"sources_read":null,"updated":"2026-10-04","human_review":null,"version":{"date":"2026-10-07"}},"gaps":[{"model_id":"gpt-6-astra","name":"GPT-6 Astra","cutoff":"2026-04","days_after":118,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":57,"in_training_data":false},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":148,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":87,"in_training_data":false}],"short_url":null}