Atlas

Models

← All models

Qwen3.8 Flash

Alibaba · 2026-08-28

Alibaba Cloud's hosted multimodal Qwen3.8 Flash model, released August 28, 2026. The official model catalog documents text, image and video input, text output, a million-token context, tool calling and optional thinking. This hosted endpoint is distinct from the separately published Qwen3.8-Flash-Next weights.

Benchmark scores

BenchmarkScore
A-Fantasia (Valid Attempts with Retries)71.7
A-Fantasia Chess (Valid Attempts with Retries)44.0
A-Fantasia Cube (Valid Attempts with Retries)72.0
A-Fantasia Spelling (Valid Attempts with Retries)99.0
BuseyBench SVG5.6
CadQueryEval (Corrected Geometry)48.0
CadQueryEval Corrected Geometry Bounding Box60.0
CadQueryEval Corrected Geometry Chamfer64.0
CadQueryEval Corrected Geometry Code Execution76.0
CadQueryEval Corrected Geometry Components72.0
CadQueryEval Corrected Geometry Hausdorff52.0
CadQueryEval Corrected Geometry STL Creation76.0
CadQueryEval Corrected Geometry Volume48.0
CadQueryEval Corrected Geometry Watertightness68.0
Capability155.0
Eval Connections One-Shot Group Accuracy95.0
Eval Connections One-Shot Puzzle Solves95.0
Eval Connections One-Shot Trap Bonus45.0
Eval Connections One-Shot with Traps75.0
GBENCH One-shot Coding GScore41.0
Humanity's Sixth Sense30.9
ObviousBench95.1
Roboflow Vision Evals — Counting, Judged (September)68.0
Roboflow Vision Evals — Counting, Strict (September)68.0
Roboflow Vision Evals — Detection mAP@50 (September)67.0
Roboflow Vision Evals — Detection mAP@50:95 (September)51.9
Roboflow Vision Evals — Detection mAP@75 (September)53.8
Roboflow Vision Evals — Extraction, Judged (September)84.9
Roboflow Vision Evals — Extraction, Strict (September)84.2
Roboflow Vision Evals — Identification, Judged (September)88.5
Loading Atlas data…