📢 Exclusive on Gate Square — #PROVE Creative Contest# is Now Live!
CandyDrop × Succinct (PROVE) — Trade to share 200,000 PROVE 👉 https://www.gate.com/announcements/article/46469
Futures Lucky Draw Challenge: Guaranteed 1 PROVE Airdrop per User 👉 https://www.gate.com/announcements/article/46491
🎁 Endless creativity · Rewards keep coming — Post to share 300 PROVE!
📅 Event PeriodAugust 12, 2025, 04:00 – August 17, 2025, 16:00 UTC
📌 How to Participate
1.Publish original content on Gate Square related to PROVE or the above activities (minimum 100 words; any format: analysis, tutorial, creativ
Google Open Source Gemma-3: Comparable to DeepSeek, Computing Power Plummets
Jinshi data news on March 13th, last night, Google (GOOG.O) CEO Sundar Pichai announced that the latest multimodal large model GEMMA-3 open source, featuring low cost and high performance. Gemma-3 has four sets of parameters: 1 billion, 4 billion, 12 billion, and 27 billion. Even with the largest 27 billion parameters, only one H100 is needed for efficient inference, which is at least 10 times more Computing Power efficient than similar models to achieve this effect, making it the currently most powerful small parameter model. According to blind test LMSYS ChatbotArena data, Gemma-3 ranks second only to DeepSeek's R1-671B, higher than OpenAI's o3-mini, Llama3-405B, and other well-known models.