Up to date
Qwen Qwen License (qwen-community-1.0) out 2 months ago
180B parameters (10 of 512 experts per token), images and text in, text out, 256K tokens of context. 2 providers serve it from $0.15 per million input tokens.
Verified 2026-10-25 Rank #38 in LLMs · score 111 method v1 Sources 4 Downloads 30d 3M · #46 of 619 tracked
Changed this week: 80 new quantized builds on the Hub (504 in all).2026-10-24 · 77 new quantized builds on the Hub (501 in all).2026-10-24
Picked by rule from 37 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Qwen3.8-Flash-Next is a 125B-parameter multimodal MoE model with 51B N-gram embeddings and 6B activated parameters per token.
Strata runs the 125-billion-parameter Qwen3.8-Flash-Next model on consumer hardware under an MIT license with no cloud dependencies.
Slotstream runs the 125-billion-parameter Qwen3.8-Flash-Next model on Macs with 16 to 64 GB of memory by streaming data from the SSD.
The 4 picks above, plus these 21. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 2 API providers; heavily run locally (504 quantized builds); a big base for builders (103 fine-tunes, adapters and merges).
Top Spaces running it: victor/Qwen3.8-Flash-Next-free-endpoint ♥ 28 · apathy-exe/Qwen3.8-Flash-Next ♥ 15 · finerobotics/annotation-studio ♥ 4 · bored10/qwen3-8-flash-next-demo ♥ 2
Weights only, an estimate: leave room for the KV cache. Method
About 108 GB at INT4, 216 GB at FP8, 432 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-08-24), amber triangles are 10 dated articles (each links to it), grey ticks are repo commits.
607 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported DeepSWE · Write code with agents | 58.7 | Agentic coding | model author | 2026-10-07 |
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Listed in LLMs
| Pipeline tag | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | Qwen4ExpForConditionalGeneration |
| Model type | qwen4_exp |
| Dtype | bfloat16 |
| Params by dtype | BF16: 179,999,981,424 · I64: 35 |
| Size band | XXL |
| MoE | yes |
| Experts | 512 |
| Experts per token | 10 |
| Context length | 262,144 (config.json max_position_embeddings) |
| License | Qwen License (qwen-community-1.0) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen4ExpForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen4ExpForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for Qwen3.8-Flash-Next: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 432 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 216 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 108 GB · Multi-GPU node (<640 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.
quantized: 504finetune: 94merge: 7adapter: 2
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF | quantized | GGUF | — | — | apache-2.0 | ISTA-DASLab | 1k | 7M |
| unsloth/Qwen3.8-Flash-Next-GGUF | quantized | GGUF | — | — | other | unsloth | 1.3k | 2.9M |
| orcarouter/Qwen3.8-Flash-Next-Uncensored-GGUF | quantized | GGUF | — | — | apache-2.0 | orcarouter | 1.1k | 1.2M |
| AtomicChat/Qwen3.8-Flash-Next-GGUF | quantized | GGUF | — | — | other | AtomicChat | 197 | 1M |
| ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-Coder-GGUF | quantized | GGUF | — | — | apache-2.0 | ISTA-DASLab | 485 | 972.8k |
| RadixArk/Qwen3.8-Flash-Next-NVFP4 | quantized | NVFP4 | 120B | XXL | other | RadixArk | 140 | 850.3k |
| Qwen/Qwen3.8-Flash-Next-FP8 | quantized | -FP8 | 180B | XXL | other | Qwen | 276 | 705.8k |
| nvidia/Qwen3.8-Flash-Next-NVFP4 | quantized | NVFP4 | 120B | XXL | other | nvidia | 387 | 588.5k |
| huihui-ai/Huihui-Qwen3.8-Flash-Next-abliterated-GGUF | quantized | GGUF | — | — | other | huihui-ai | 186 | 458.7k |
| Maximilian228/Qwen3.8-Flash-Next-NVFP4-GPTQ-Strata | quantized | NVFP4 | — | — | other | Maximilian228 | 37 | 244.9k |
| agentionai/Qwen3.8-Flash-Next-Gyro-GGUF | quantized | GGUF | — | — | other | agentionai | 115 | 229.9k |
| antirez/qwen3.8-flash-next-gguf | quantized | gguf | — | — | other | antirez | 14 | 230.5k |
| bartowski/Qwen3.8-Flash-Next-GGUF | quantized | GGUF | — | — | other | bartowski | 44 | 154.7k |
| local-inference-lab/Qwen3.8-Flash-Next-NVFP4 | quantized | NVFP4 | 92.7B | XL | other | local-inference-lab | 56 | 138.2k |
| huihui-ai/Huihui-Qwen3.8-Flash-Next-abliterated-v2-GGUF | quantized | GGUF | — | — | other | huihui-ai | 122 | 129.9k |
| pentacoxian-dev/Qwen3.8-Flash-Next-IQ3E-Q8D-MTP-GGUF | quantized | GGUF | — | — | other | pentacoxian-dev | 21 | 130.2k |
| Inferact/Qwen3.8-Flash-Next-NVFP4 | quantized | NVFP4 | 118B | XXL | other | Inferact | 30 | 120.3k |
| agentionai/Qwen3.8-Flash-Next-ROCmFP4-FAST-imatrix-GGUF | quantized | imatrix | — | — | other | agentionai | 88 | 109.9k |
| orcarouter/Qwen3.8-Flash-Next-Uncensored-FP8 | quantized | -FP8 | 180B | XXL | apache-2.0 | orcarouter | 57 | 110.6k |
| sh0wie/Qwen3.8-Flash-Next-REAP-288-GGUF | quantized | GGUF | — | — | other | sh0wie | 11 | 112.7k |
| ggml-org/Qwen3.8-Flash-Next-GGUF | quantized | GGUF | — | — | other | ggml-org | 36 | 72.2k |
| ivanfioravanti/Qwen3.8-Flash-Next-DS4-Q4 | quantized | GGUF | — | — | other | ivanfioravanti | 16 | 70.5k |
| AnonimousA/Qwen3.8-Flash-Next-REAP-256-duo-GGUF | quantized | GGUF | — | — | other | AnonimousA | 21 | 68.9k |
| Intel/Qwen3.8-Flash-Next-W4A16-AutoRound | quantized | -W4A16 | 75.4B | XL | other | Intel | 36 | 66.8k |
| agentionai/Qwen3.8-Flash-Next-AP-GGUF | quantized | GGUF | — | — | other | agentionai | 37 | 64.5k |
| pfeifferj/Qwen3.8-Flash-Next-GSQ-RCO-GGUF | quantized | GGUF | — | — | other | pfeifferj | 51 | 63.1k |
| Jundot/Qwen3.8-Flash-Next-oQ4e-mtp | quantized | — | 180B | XXL | other | Jundot | 103 | 59.9k |
| primitive-ai/Qwen3.8-Flash-Next-NVFP4 | quantized | NVFP4 | 120B | XXL | other | primitive-ai | 25 | 62.5k |
| kingjones777/Qwen3.8-Flash-Next-Uncensored-ROCmFP4-FAST-GGUF | quantized | GGUF | — | — | other | kingjones777 | 9 | 56.4k |
| sh0wie/Qwen3.8-Flash-Next-REAP-288-MLX-4bit | quantized | MLX | 125B | XXL | other | sh0wie | 64 | 45.5k |
| jcbtc/Qwen3.8-Flash-CIRU-STRIX-IU4 | quantized | GGUF | — | — | other | jcbtc | 41 | 41.6k |
| agentionai/Signal-3.8-Flash-Next-GGUF | finetune | GGUF | — | — | other | agentionai | 35 | 40.8k |
| AnonimousA/Qwen3.8-Flash-Next-REAP-320-GGUF | quantized | GGUF | — | — | other | AnonimousA | 25 | 37.6k |
| Youssofal/Qwen3.8-Flash-Next-MTPLX-Optimized-Speed | quantized | — | 126B | XXL | other | Youssofal | 44 | 35.9k |
| jcbtc/Qwen3.8-Flash-CIRU-STRIX-Orca | quantized | GGUF | — | — | other | jcbtc | 40 | 36.1k |
| nitinpanj/qwen38-flash-next-v3 | quantized | GGUF | — | — | — | nitinpanj | 5 | 36.5k |
| lmstudio-community/Qwen3.8-Flash-Next-GGUF | quantized | GGUF | — | — | other | lmstudio-community | 6 | 36.3k |
| ivanfioravanti/Qwen3.8-Flash-Next-DS4-IQ2 | quantized | GGUF | — | — | other | ivanfioravanti | 20 | 32.7k |
| Sawfwair/Qwen3.8-Flash-Next-MLX-Mixed-2bit | quantized | MLX | 180B | XXL | other | Sawfwair | 12 | 31.2k |
| nvidia/Qwen3.8-Flash-Next-FP4-Local | quantized | — | 96.5B | XL | other | nvidia | 329 | 13.8k |
Hosted endpoints observed for Qwen3.8-Flash-Next, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.15/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Alibaba openrouter | qwen/qwen3.8-flash | 977K | unknown | $0.15 | $0.47 | — | — | yes | live | 2026-10-24 |
| Featherless hf | Qwen/Qwen3.8-Flash-Next | — | — | — | — | — | — | — | live | 2026-10-25 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:32 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
