Expired50% confidenceFactExact time
DeepSeek V3 introduces an FP8 mixed-precision training framework during the pre-training phase, and its API pricing is approximately 1/20 to 1/10 that of GPT-4o.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/13/2026
First Seen
Valid until: 10/11/2026(expired)
Sources
一句话生成小游戏？Cocos AI智能体插件实测解析
bilibili起飞的鳗鱼7/11/2026
Related Claims
VerifiedDeepSeek V3 adopts a MoE architecture, reducing training and inference costs to a fraction of those of GPT-4-level models while maintaining top-tier performance.82% similarUnverifiedDeepSeek-V3 and the Qwen2.5 series have achieved performance approaching GPT-4 level on code and structured data generation benchmarks (such as HumanEval and MBPP), while their API call costs are only about 1/10 of GPT-4's.77% similarUnverifiedThe training cost of DeepSeek-V3 is only one-twentieth that of GPT.75% similarVerifiedDeepSeek's R1 model has demonstrated reasoning capabilities comparable to GPT-4o across multiple benchmarks, and its training cost is reportedly only a tiny fraction of that of comparable American models.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/501997API
curl https://kongchang.com/api/v1/knowledge/claims/501997MCP
get_claim(id=501997)