Qwen / Qwen3-VL 8B Thinking

BF16 quantization for Qwen3-VL 8B Thinking

The published profile is indicative; engineering measurements are completed for the selected model and hardware.

Before → After

Base releaseVerified checkpoint
8.000B

QuantizationBF16
Quality retention target

Production artifactArtifact
Lineage and benchmark report

Weights and KV cache

Estimated until measured during engineering

Estimated VRAM

Measured and certified results are delivered with the project.

Runtime packaging

transformers

Not currently orderable