Unverified50% confidenceBenchmarkExact time
Kimi-K2.7-Code can be trained on 6 B200 GPUs with packed INT4 experts, whereas using 16-bit copies would require approximately 2 TB of VRAM.
1
Sources
50%
Confidence
Long-term
Relevance
10/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedKimi K2在FP16精度下仅模型权重就需要数百GB显存，INT4量化可将显存需求压缩至约四分之一70% similarUnverifiedThe Kimi K2 series supports a 128K token context.65% similarUnverifiedThe test used the Swift version's Q5_K_L and the base version's Unsloth Q5_K_XL quantization configurations, both running in 128GB of memory with support for a 262K context, with engrams in Q8_0 stored on SSD.64% similarUnverifiedKimi K3 requires approximately 1.5 TB of memory under Q4 quantization, potentially meaning four Macs would need to be linked together.63% similarVerifiedFor a typical 7B model with 32 layers and 32 attention heads, storing the KV Cache for each token requires approximately 0.5 MB of VRAM, so generating 1,000 tokens requires an additional ~500 MB of VRAM62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/965336API
curl https://kongchang.com/api/v1/knowledge/claims/965336MCP
get_claim(id=965336)