Expired50% confidenceFactExact time
TPUv4 has already supported the training of large models such as Gemini.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/12/2026
First Seen
Valid until: 10/10/2026(expired)
Sources
DeepSeek Builds Its Own AI Chips: The Strategic Logic and Challenges Behind Compute Independence
hackernewshackernews7/9/2026
Related Claims
VerifiedThe Gemini series jointly models multiple modalities from the training stage, rather than appending a visual encoder as solutions like GPT-4V do.78% similarUnverified谷歌于2016年正式发布第一代张量处理单元（TPU），核心是脉动阵列架构，目前TPUv4已支撑Gemini等大模型训练75% similarUnverifiedGoogle自研的TPU芯片为Gemini模型的训练和推理提供定制化算力支持，在成本和效率上相比依赖英伟达GPU的竞争对手具备一定优势74% similarUnverifiedThe training of the Gemini series of models is based on JAX.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/489225API
curl https://kongchang.com/api/v1/knowledge/claims/489225MCP
get_claim(id=489225)