Qwen3.8 Flash

Available as qwen3.8-flash through Qwen Cloud; the underlying Qwen3.8-Flash-Next weights are downloadable

Publisher-reported capabilities

• Qwen3.8 Flash is a multimodal reasoning model with a 1M-token hosted context window and up to 131k output tokens.
• The open Qwen3.8-Flash-Next checkpoint uses a hybrid architecture with 125B language-model parameters, 6B active parameters, and large n-gram embeddings.

Limitations and availability caveats

• The downloadable Flash-Next checkpoint has a native 262k-token context; reaching 1M locally requires RoPE scaling that can affect shorter-context performance.
• Provider-reported Flash-Next benchmarks are not copied into AA- or Vals-owned sparse fields until those canonical sources publish compatible rows.

Official release and model references

• qwen.ai/blog
• www.qwencloud.com/models/qwen3.8-flash
• huggingface.co/Qwen/Qwen3.8-Flash-Next

Qwen3.8 Flash has 12 of 39 tracked IQ benchmark inputs across 4 of 7 dimensions. Raw benchmark tables contain source-backed results only. Derived IQ combines all six dimensions using the shared scoring methodology, including estimates where direct coverage is incomplete. Configured predecessor references: Qwen3.7 Flash. A reference is an estimation input, not evidence that this model has the predecessor's measured results.

Among dimensions with direct benchmark coverage, this profile's highest estimated scores are Programmatic Reasoning (IQ 136) and Academic Reasoning (IQ 129). These dimension estimates can include statistical adjustments or predecessor priors; they are not raw benchmark percentages or claims of real-world task success.

Benchmark versions, evaluation harnesses, and coverage affect comparisons; a missing result is not a score of zero. See the scoring methodology for estimation and uncertainty limits.