Unverified50% confidenceFactExact time
Teams such as DeepSeek and Qwen, when releasing important models, follow a cadence of making the API available first for days to weeks before releasing the weights.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/5/2026
First Seen
Valid until: 11/3/2026
Sources
OpenRouter聚合平台迎新模型，开放权重即将发布意味着什么
Related Claims
UnverifiedDeepSeek simultaneously open-sourced a complete project named DeepSpark, including training, evaluation, and experiment code, which can be ported to mainstream models such as Qwen and Gemma73% similarUnverifiedDomestic models such as DeepSeek-V3 and Qwen have achieved a considerable degree of performance replication through model distillation techniques, with parameter counts only a fraction of OpenAI's top models.68% similarUnverifiedBoth DeepSeek-R1 and the Qwen series of models have publicly discussed the effectiveness of the code-as-reward training paradigm.68% similarUnverified使用DeepSpec的训练流水线，团队可在两周内完成自家模型的推测解码适配67% similarVerifiedDeepSeek V4 lags behind top closed-source models in knowledge reasoning by approximately three to six months.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/686974API
curl https://kongchang.com/api/v1/knowledge/claims/686974MCP
get_claim(id=686974)