Unverified50% confidenceEventExact time
Developer Hugo Vergnes trained a 3.8-billion-parameter large language model to a performance level of 0.384 CORE for less than $1,000 (USD 998).
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA developer following Karpathy's nanochat project trained 9 checkpoints across 3 different versions of LLMs from scratch with a budget of approximately $750, with parameter counts ranging from 353M to 672M.69% similarUnverifiedR1的基座模型V3训练成本仅558万美元，后续强化学习部分花费29万美元65% similarUnverified基于开源模型使用LoRA、QLoRA等微调方法的成本可低至数万美元，远低于从零训练所需的上亿美元64% similarUnverifiedSmoLLM的训练总成本不到人民币350元（约合4000卢比）63% similarUnverified当前主流LLM（GPT-4、Claude、Gemini、Llama等）参数量从数十亿到数万亿不等，训练成本从数百万到数千万美元63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/892928API
curl https://kongchang.com/api/v1/knowledge/claims/892928MCP
get_claim(id=892928)