Unverified50% confidenceBenchmarkExact time
The 4-billion-parameter version of Qwen 3.5, with no external tools and with exact calculation required at each step, got only 22 of 40 multi-digit multiplication problems correct.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/31/2026
First Seen
Valid until: 1/29/2027
Sources
Related Entities
Related Claims
UnverifiedTongyi Qianwen 3.8 has approximately 2 trillion parameters, and Kimi has approximately 2 trillion parameters.78% similarUnverifiedFor a configuration of 1.6 trillion total parameters and 49 billion activated parameters, only about 3% of parameters participate in computation during actual inference.77% similarUnverifiedA model with 100 billion parameters requires approximately 80 GB of VRAM to process a single inference request, while a model with tens of billions of parameters requires only about 20 GB.75% similarUnverifiedThe full Neural 3 model with 121 million parameters has an actual computational load roughly equivalent to a 50-million-parameter model.75% similarUnverifiedTaking a 1000×1000 weight matrix as an example, if the LoRA rank is set to 8, only approximately 16,000 parameters need to be trained instead of the original 1,000,000 parameters.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1121140API
curl https://kongchang.com/api/v1/knowledge/claims/1121140MCP
get_claim(id=1121140)