Expired50% confidenceBenchmarkExact time
In Nodus's pilot on October 11, 2026, using a Qwen2.5-0.5B model running the same recipe on a single NVIDIA L4, supervised fine-tuning cost 67% less than Modal and continued pretraining cost 42.5% less; after counting one failed attempt, the fine-tuning savings dropped to 24.7%.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/12/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/26/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedTransferring the PLE (n-gram memory) pre-trained by Qwen3.8 Flash-Next to the Qwen3.5-0.8B model achieved a 5.05% perplexity reduction on the frozen validation set.68% similarUnverifiedQwen2.5系列在后训练阶段引入了大量工具调用对齐数据，并通过RLHF和DPO强化了结构化输出的可靠性65% similarUnverifiedQwen3-0.8B is a small-parameter model that can be trained on a single consumer-grade GPU.65% similarUnverifiedQwen-Image-2.1-Turbo maintains the same 7B parameter architecture as the base model, with the number of denoising steps reduced by 5×63% similarUnverified第一代Qwen起步上下文窗口为2048，通过NTK插值外推到8192，训练使用Flash Attention、AdamW优化器和BF16混合精度61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1007359API
curl https://kongchang.com/api/v1/knowledge/claims/1007359MCP
get_claim(id=1007359)