Unverified50% confidenceBenchmarkExact time
In the front-end Bund scenery test, 4.1 Flash consumed twice as many tokens as the previous generation.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
UnverifiedFlashAttention在长序列场景下速度提升达2-4倍64% similarUnverifiedFlashVision EXP supports a context length of up to 1 million tokens.60% similarUnverifiedThe 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.56% similarUnverifiedFlash Attention was proposed by Tri Dao and achieves a 2–4× speedup in practice.55% similarUnverifiedFlashVision EXP has a single-image input limit of 384 tokens.54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/908312API
curl https://kongchang.com/api/v1/knowledge/claims/908312MCP
get_claim(id=908312)