Experimental vision-enabled DeepSeek V4 Flash with image understanding added on top of the 0731 snapshot.
Add a second model to compare.