Unverified80% confidenceEventTime unknown
百度千帆发布端到端文档智能模型千帆OCR，参数量4B，在多项基准测试中比DeepSeek OCR2略高，也领先Gemini 3 Pro
1
Sources
80%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026
Sources
AI日报3月20日：Gemini CLI限制免费用户，小米MiMo三模型齐发
bilibiliinfinite灵感港
Related Entities
Related Claims
Unverified在OmniDocBench V1.5测试中，OCR2在使用与Gemini 3 Pro相同数量的token的情况下取得了更高的性能表现72% similarUnverifiedGoogle Gemini在最新的数学推理和编程基准测试中已经超越了OpenAI O3、Claude 4和DeepSeek R1，稳居榜首71% similarUnverifiedGemini 3在SWE-bench基准测试中智能体场景下的得分远超同类模型70% similarUnverifiedGemini 2.5 Pro在数学能力基准测试中与O3 Mini持平，两者都将其他模型远远甩在身后69% similarUnverified代码中还提到了一个名为'King4'的模型，据称在早期测试中表现出强劲性能，尚不清楚是Gemini 3的早期版本还是Gemini 2.5 Pro搭载Deep Think功能的变体68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/22827API
curl https://kongchang.com/api/v1/knowledge/claims/22827MCP
get_claim(id=22827)