Unverified50% confidenceFactExact time
The FlashVision EXP model has a total parameter count of 304B, uses a MoE architecture, activates only 6 experts during inference, and employs FP4 quantization precision.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFlashVision EXP has a single-image input limit of 384 tokens.70% similarUnverifiedFlashVision EXP supports a context length of up to 1 million tokens.69% similarUnverifiedIn the Artificial Analysis benchmark, GLM 5.3 Flash has a hallucination score of 7 and Qwen 3.8 (27B) scores -10, with local models having higher hallucination rates than cloud-based ones.65% similarUnverified4K resolution output means the model needs to maintain detail consistency at the 4096×4096 pixel level.61% similarUnverifiedIn multimodal Agent evaluations, the capabilities of DeepSeek-V-Flash-Vision-XP approach those of Claude Opus 4.8.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/904966API
curl https://kongchang.com/api/v1/knowledge/claims/904966MCP
get_claim(id=904966)