estimate, except measured file sizes

Can I run Qwen3.8-Flash-Next locally?

180B total, 6B active. Parameter count is an estimate.

Licence:

Qwen Community License 1.0 source

Context:

262,144 tokens

Modality:

text + image/video in, text out

Released:

not recorded

Mixture-of-experts vision-language model. Hugging Face metadata records license: other and license_name: qwen-community-1.0. The LICENSE file is titled Qwen Community License 1.0. It requires the copyright notice. A commercial product or service with more than 100 million monthly active users, or more than US$20 million in monthly revenue, must display the model name. A Model as a Service or AI Work Assistant business needs a separate licence from Qwen before commercial use. Internal use that does not make the model or its outputs available to a third party is exempt from that separate licence. The card says 125B with 6B activated, plus 51B n-gram embedding and 4B MTP. Those round billions are why paramsEstimated stays true. Hugging Face safetensors metadata lists 179,999,981,459 parameters (BF16 179,999,981,424 and I64 35). VRAM math uses that total, stored as 180B. 6B active is the card's activated figure for the 125B portion, not a second measurement. 512 experts in the config, 10 routed per token. The card adds 1 shared expert. 48 layers: 36 linear_attention and 12 full_attention. Those 12 are Qwen Sparse Attention. The card's budget is 512 blocks or 2,048 tokens, and text_config.indexer_budget is 2048. That cache is outside this formula, so kvNotModelled is set and the KV column is not a computed cache. 2 KV heads, head dim 256, max_position_embeddings 262,144. The card says the window is extensible to 1,000,000 tokens with a RoPE change; this catalog uses 262,144. The card's citation is August 2026. Hugging Face lists the repo as created on 24 August 2026. Speed is not estimated: the formula streams every weight and understates sparse MoE, and no speed figure is stored. Unsloth GGUF sizes, shards summed, decimal GB, checked 19 Oct 2026: UD-IQ4_XS 93.68, UD-Q4_K_XL 111.33, Q8_0 188.23. MTP/mtp-Qwen3.8-Flash-Next-Q8_0.gguf is 4.14 GB (4,137,429,120 bytes) and is not included in those three totals. mmproj-F16 is 0.90 GB (904,004,000 bytes), also separate. Strata's README says the graphics card keeps the experts that are used most often and RAM holds all of them (https://github.com/Niko1221/Strata). Releases: https://github.com/Niko1221/Strata/releases. Mainline llama.cpp GPU expert-cache pull requests were still open on 19 Oct 2026: https://github.com/ggml-org/llama.cpp/pull/27861, https://github.com/ggml-org/llama.cpp/pull/30594 and https://github.com/ggml-org/llama.cpp/pull/30637.

Parameter source · Architecture source · Model page

speed: not estimated for MoE yet

KV cache not modelled for this architecture: real use needs more than the figures shown

Memory per quant

KV is at 8,192 tokens. Overhead is 0.8 GB. Both are estimates. A measured file size is labelled. The last column is the cheapest machine that fits the total fully.

Cheapest rig that fits

UD-Q4_K_XL, full: Mac Studio M5 Ultra 256GB, €12,500 (estimate)

UD-Q4_K_XL, slow, partly in system RAM: Dual RTX 3090, 48 GB, slow, partly in system RAM, €5,408 (estimate)

UD-IQ4_XS (smallest file), full: Mac Studio M4 Max 128GB, €3,290 (estimate)

UD-IQ4_XS (smallest file), slow, partly in system RAM: Dual RTX 3090, 48 GB, slow, partly in system RAM, €5,408 (estimate)

Prices moved: see Building a local LLM rig in the RAMageddon.

On a specific GPU

• Qwen3.8-Flash-Next on RTX 3060 12GB
• Qwen3.8-Flash-Next on RTX 4060 Ti 16GB
• Qwen3.8-Flash-Next on RTX 3090
• Qwen3.8-Flash-Next on RTX 4090
• Qwen3.8-Flash-Next on RTX 5080
• Qwen3.8-Flash-Next on RTX 5090
• Qwen3.8-Flash-Next on RX 7900 XTX
• Qwen3.8-Flash-Next on Dual RTX 3090
• Qwen3.8-Flash-Next on Dual RTX 4090