Unverified60% confidenceOpinionExact time
Optimizing prompts costs almost nothing and is the most cost-effective way to improve model performance, whereas training a model from scratch may cost tens of millions.
2
Sources
60%
Confidence
Long-term
Relevance
10/24/2026
First Seen
Sources
Related Entities
Related Claims
Unverified训练2.8万亿参数模型代价高昂，便宜和开源并不等于不需要算力，算力需求从闭源巨头向分散的企业定制场景转移78% similarUnverifiedTraining a frontier model can cost tens of millions to hundreds of millions of dollars, and once its API outputs are harvested at scale, competitors can train substitutes with comparable capabilities at a cost far below the training cost, directly eroding the original model vendor's commercial moat.75% similarUnverifiedReasoning models are extremely costly to train, typically requiring a combination of large-scale reinforcement learning and carefully designed reward models to acquire the capability.74% similarUnverifiedOnly when a certain type of task is sufficiently high-frequency, stable, and cost-sensitive does migrating it from a general-purpose model to a specially trained model potentially realize significant cost benefits.74% similarUnverifiedModel distillation allows attackers to obtain near-equivalent capabilities without bearing the multi-million-dollar compute costs of original pre-training, requiring only modest API call fees and fine-tuning compute.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1092415API
curl https://kongchang.com/api/v1/knowledge/claims/1092415MCP
get_claim(id=1092415)