Unverified50% confidenceBenchmarkExact time
On two existing large models, DeepSeek-V4.1-Flash and Qwen3.8-Flash-Next, the low-bit tables saved 47% to 55% of table memory.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/23/2026
First Seen
Valid until: 1/21/2027
Sources
Related Entities
Related Claims
UnverifiedThe N-Gram Embedding parameters of Qwen3.8-Flash-Next account for approximately 51B and can be offloaded to CPU memory to reduce GPU VRAM requirements.70% similarUnverifiedFlash 1017 scores 68.9% on LongBench QA top-5, higher than DeepSeek V4-Pro's 66.1% and GPT-4-Turbo's 62.3%.69% similarUnverifiedThere is no all-around winner among the three lightweight models: DeepSeek V4.1 Flash has the highest quality but is slow, Haiku 5.5 is fast and leads in writing but lags in programming visual output, and GPT-6 Luna performed poorly on tasks.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1088229API
curl https://kongchang.com/api/v1/knowledge/claims/1088229MCP
get_claim(id=1088229)