Web 3.0 Ethiopia - DeFi & AI – Telegram
Web 3.0 Ethiopia - DeFi & AI
602 subscribers
718 photos
14 videos
5 files
149 links
Bridging the Information Gap on DeFI and Artificial Intelligence for Ethiopians
Download Telegram
GPT-5.2-Codex Sets New Benchmarks in Agentic Coding

GPT-5.2-Codex sets new benchmarks in agentic coding with 56.4% accuracy on SWE-Bench Pro and 64.0% on Terminal-Bench 2.0, improving over GPT-5.2 by 0.8% and 1.8% respectively, enabling reliable multi-file refactors and long-context tasks via native compaction. Now accessible in Codex for paid ChatGPT users with API rollout imminent.

ጂፒቲ-5.2-ኮዴክስ (GPT-5.2-Codex) በወኪል ተኮር የኮዲንግ ስራ (agentic coding) ላይ አዲስ የብቃት ደረጃን ያስመዘገበ ሲሆን፣ በ SWE-Bench Pro ላይ 56.4% እንዲሁም በ Terminal-Bench 2.0 ላይ 64.0% ትክክለኛነትን አሳይቷል። ይህም ከቀደመው ጂፒቲ-5.2 (GPT-5.2) ጋር ሲነጻጸር በቅደም ተከተል የ0.8% እና የ1.8% መሻሻልን ያሳያል። ይህ ሞዴል "native compaction" በተሰኘ ቴክኖሎጂ አማካኝነት በርካታ ፋይሎችን በአንድ ጊዜ የማስተካከል (multi-file refactors) እና ረጅም ይዘት ያላቸውን ውስብስብ ስራዎች በላቀ አስተማማኝነት ማከናወን ያስችላል። አገልግሎቱ አሁን ላይ ለቻትጂፒቲ (ChatGPT) ከፍለው ለሚጠቀሙ ደንበኞች በኮዴክስ (Codex) በኩል ክፍት የሆነ ሲሆን፣ ኤፒአይ (API) አገልግሎቱም በቅርቡ ይለቀቃል።

@webthreeth
3
Minimax releases Minimax M2.1

MiniMax M2.1, an open-source upgrade to the October 2025 M2 model, now integrated into the MiniMax Agent platform for enhanced coding and automation. Core improvements focus on multilingual code handling for tasks like test generation and optimization, robust agentic tool chains for browser-based workflows, and "Digital Employee" features for multi-step office processes.

የጥቅምት 2025ቱ "ኤም2" (M2) ሞዴል ክፍት (open-source) ማሻሻያ የሆነው "ሚኒማክስ ኤም2.1" (MiniMax M2.1)፣ ለተሻሻለ የኮዲንግ እና አውቶሜሽን አቅም ወደ "ሚኒማክስ ኤጀንት ፕላትፎርም" (MiniMax Agent platform) ተቀላቅሏል። የዚህ ሞዴል ዋና ዋና ማሻሻያዎች የሚያተኩሩት እንደ የሙከራ ፈጠራ (test generation) እና አሰራርን ማመቻቸት (optimization) ላሉ ተግባራት በሚያገለግለው የብዝሃ-ቋንቋ የኮድ አያያዝ (multilingual code handling) ላይ ነው። በተጨማሪም፣ በአሳሽ/ብራውዘር ለሚከናወኑ ስራዎች አስተማማኝ የወኪል መሳሪያዎች (agentic tool chains) እና ለተወሳሰቡ የቢሮ ሂደቶች የሚረዱ "ዲጂታል ሰራተኛ" (Digital Employee) ባህሪያትን አካቷል።

@webthreeth
Zhipu AI Unveils GLM-4.7: A New Frontier in Open-Source Reasoning and Coding

Zhipu AI has launched GLM-4.7, a state-of-the-art open-source model that achieves top scores in coding and reasoning, rivaling elite proprietary systems like Claude 4.5 Sonnet. The model introduces refined thinking modes for complex multi-step tasks and features advanced capabilities in frontend optimization and aesthetic design for professional presentations. Targeting developers and enterprises, GLM-4.7 offers expanded context handling and improved instruction following through Z.ai’s platform and APIs.

ዚፑ ኤአይ (Zhipu AI) በኮዲንግ እና በምክንያታዊ ትንተና ከፍተኛ ውጤት ያስመዘገበውንና እንደ ክሎድ 4.5 ሶኔት (Claude 4.5 Sonnet) ካሉ የግል ሞዴሎች ጋር የሚወዳደረውን GLM-4.7 የተሰኘ አዲስ የክፍት ምንጭ ሞዴል ይፋ አድርጓል። ሞዴሉ ለውስብስብ ባለብዙ-ደረጃ ተግባራት የተሻሻሉ የአስተሳሰብ ዘዴዎችን ያካተተ ሲሆን፣ ለድረ-ገጽ ማሻሻያ እና ለተለያዩ ዲዛይኖች የላቀ ብቃትን ያሳያል። ይህ ለሶፍትዌር አልሚዎች እና ለድርጅቶች የታለመው ሞዴል ሰፊ መረጃዎችን የማስተናገድ እና ትዕዛዞችን የመከተል ችሎታውን በዚ.ኤአይ (Z.ai) መድረክ እና ኤፒአይ (API) በኩል ያቀርባል።

@webthreeth
Liquid AI’s (3B) Open Model Edges Out GPT-4o and DeepSeek in Recent Tests

Liquid AI unveiled an experimental open model it says can run locally, and posted results that beat bigger rivals on key tests. On instruction-following, the company’s reported score comes in above OpenAI’s GPT-4o on the same measure, and it also edges DeepSeek’s R1-0528 on a separate instruction benchmark. For the industry, it’s another sign of how much progress smaller AI systems have made in just the past year.

ሊኩዊድ ኤአይ (Liquid AI) በራስ ኮምፒውተር (locally) ላይ ሊሰራ የሚችልና በቁልፍ ፈተናዎች ላይ ትልልቅ ተፎካካሪዎቹን ያሸነፈ ሙከራ ላይ ያለ ግልጽ (open) ሞዴል ይፋ አድርጓል። መመሪያዎችን በመከተል ረገድ የኩባንያው ውጤት በዚሁ መመዘኛ ከኦፕን ኤአይ (OpenAI) ጂፒቲ-4ኦ (GPT-4o) በላይ ሆኖ ተመዝግቧል፤ እንዲሁም በሌላ የመመሪያ መመዘኛ ላይ የዲፕሲክን (DeepSeek) አር1-0528 (R1-0528) በትንሽ ልዩነት በልጧል። ይህም ለዘርፉ ትናንሽ የሰው ሰራሽ አስተውሎት (AI) ሥርዓቶች ባለፈው አንድ ዓመት ውስጥ ብቻ ምን ያህል ትልቅ እድገት እንዳሳዩ ሌላው ማሳያ ነው።

@webthreeth
3
Trend #1 - AI Competitive Landscape Shifted Dramatically (End of 2025 Trends)

OpenAI's lead has diminished as competitors like Gemini 3 Pro claim the top spot. Technical benchmarks show a narrowing gap, transforming the market into a balanced and highly competitive arena where rival models now match or surpass OpenAI's performance. Data from the LM-Arena rankings confirms this shift, showing Gemini 3 Pro in 1st place while GPT 5.2 has fallen to 8th. This shift indicates that technical superiority is no longer exclusive to a single developer in the rapidly advancing AI industry.

እንደ Gemini 3 Pro ያሉ ተወዳዳሪዎች ቀዳሚነቱን በመያዛቸው የOpenAI የቴክኖሎጂ የበላይነት እየቀነሰ መጥቷል። የተለያዩ የብቃት መመዘኛዎች በኩባንያዎች መካከል ያለው የቴክኒክ ልዩነት እየጠበበ መሆኑን ያሳያሉ፤ ይህም ዘርፉን ከአንድ መሪነት ወደ ተመጣጣኝና ከፍተኛ ፉክክር ያለበት መድረክ ቀይሮታል። የLM-Arena የደረጃ መረጃ Gemini 3 Pro አንደኛ ደረጃን ሲይዝ GPT 5.2 ግን ወደ ስምንተኛ ደረጃ ዝቅ ማለቱን በማሳየት ለዚህ ለውጥ በቂ ማረጋገጫ ይሰጣል። ይህም በዘርፉ ያለው የቴክኖሎጂ የበላይነት ከአንድ ኩባንያ ቁጥጥር ውጭ እየሆነ መምጣቱንና ፈጣን እድገት እየታየ መሆኑን ያረጋግጣል።

@webthreeth
4
Why are people goating Claude OPUS 4.5 like AGI? Does anyone have any idea?
🤔1
Meta Bolsters Global AI Agent Strategy with Manus AI Acquisition

Meta has acquired Manus AI—its fifth AI deal in 2025—bringing $100M in annual recurring revenue and expertise in handling 147 trillion tokens on 80 million virtual computers. The move ensures seamless service for millions of subscribers while scaling general AI agents to billions through Meta's platforms, including integrations like Meta AI for consumers and enterprises.

ሜታ (Meta) ማኑስ ኤአይን (Manus AI) በመግዛት በ2025 አምስተኛውን የአርቴፊሻል ኢንተለጀንስ (AI) ስምምነቱን ፈጽሟል፤ ይህም 100 ሚሊዮን ዶላር ዓመታዊ ተደጋጋሚ ገቢን እንዲሁም 147 ትሪሊዮን ቶከኖችን በ80 ሚሊዮን ምናባዊ ኮምፒውተሮች ላይ የማስተናገድ ብቃትን ይዞ መጥቷል። ይህ እርምጃ ለሚሊዮኖች ተመዝጋቢዎች ቀጣይነት ያለው አገልግሎት የሚያረጋግጥ ሲሆን፣ እንደ ሜታ ኤአይ ለተጠቃሚዎችና ለድርጅቶች ያሉ መድረኮችን በመጠቀም አጠቃላይ የኤአይ ወኪሎችን ለቢሊዮኖች ለማድረስ ያስችላል፤ ዝርዝሩም በኩባንያው ይፋዊ ብሎግ ላይ ተገልጿል።

@webthreeth
1
Trend #2 - AI Capabilities have had High Growth (End of 2025 Trends)

AI capabilities have had an exponential growth, with key benchmarks showing a two-to-threefold improvement in just one year. Significant milestones include Poetiq’s Meta-System achieving a 75% score on ARC-AGI-2, officially surpassing the human average of 60%. Performance also surged in other areas, as SWE-Bench scores nearly doubled to 80.9% while HLE scores saw a remarkable sixfold increase to 50%.

በኤአይ (AI) ዘርፍ በአንድ ዓመት ውስጥ የታየውን ወደር የለሽ እና እጅግ ፈጣን እድገት በዝርዝር ያሳያል። በARC-AGI-2 መመዘኛ የPoetiq Meta-System 75% በማምጣት የሰዎችን አማካይ ውጤት (60%) መብለጥ መቻሉ ትልቅ የቴክኖሎጂ ስኬት ሆኖ ተመዝግቧል። በተጨማሪም የSWE-Bench ውጤት በአንድ ዓመት ውስጥ በእጥፍ ሲያድግ፣ የHLE ውጤት ደግሞ ወደ 50% በመሸጋገር የስድስት እጥፍ ከፍተኛ ጭማሪ አሳይቷል።

@webthreeth
1
Trend #3 - AI Sector Sees Explosive Growth in Valuations and Revenue (Trends of 2025)

The AI sector is experiencing explosive growth, with the "Big Four" companies seeing their valuation surge 2,587% to $833 billion. This expansion is driven by significant revenue increases at OpenAI (up 250% to $19B) and Anthropic (up 600% to $7B), reflecting robust market confidence in the technology.

የአርቴፊሻል ኢንተለጀንስ (ኤአይ) ዘርፍ እጅግ ፈጣን እድገት እያሳየ ሲሆን፣ "አራቱ ግዙፍ" ኩባንያዎች የገበያ ዋጋ በ2,587 በመቶ በመጨመር 833 ቢሊዮን ዶላር ደርሷል። ለዚህ እድገት ዋነኛው ምክንያት የኦፕን ኤአይ (OpenAI) ገቢ በ250 በመቶ (19 ቢሊዮን ዶላር) እና የአንትሮፒክ (Anthropic) ገቢ በ600 በመቶ (7 ቢሊዮን ዶላር) መጨመሩ ሲሆን፤ ይህም ገበያው በቴክኖሎጂው ላይ ያለውን ጠንካራ እምነት የሚያንጸባርቅ ነው።

@webthreeth
1
Happy 1st Anniversary to Web 3.0 Ethiopia - DeFi & AI!

It’s hard to believe, but today marks one full year since we launched Web 3.0 Ethiopia - DeFi & AI! In the last 365 days, I've shared nearly 1,000 news pieces on AI and DeFi, all aimed at one goal: showing you how to put AI to work, whether you're in the field or in the office.

A massive thank you and shout-out to our incredible community on LinkedIn and Telegram! You are the reason this channel thrives. You've advertised the channel simply because you believe in it—without any payment. You’ve shared the benefits with everyone you know. You've given me critical feedback on how to improve and suggested ways we can partner to make a bigger difference. Everything means a lot.

You've shown me the immense power of good in this world. This channel’s success belongs entirely to its followers. Every achievement this past year is yours.

Thank you again. I am truly indebted and excited to deliver even more in 2026.
👏41
The Emergence of IQuest-Coder-V1 from China

IQuest-Coder-V1 is a groundbreaking 40B-parameter coding model from China's Quest Research that has redefined industry standards by outperforming giants like GPT-5.1 on the SWE-Bench Verified benchmark, with a score of 81.4%. Utilizing a unique recurrent Loop architecture and a 128K context window, the model features specialized "Thinking" and "Instruct" variants designed for deep reasoning and precise execution.

IQuest-Coder-V1 በቻይናው ኩዌስት ሪሰርች (Quest Research) የተገነባ 40 ቢሊዮን ፓራሜትር ያለው እጅግ ዘመናዊ የኮዲንግ ሞዴል ነው። ይህ ሞዴል በSWE-Bench Verified መለኪያ ላይ እንደ GPT-5.1 ያሉ ታላላቅ ሞዴሎችን በመብለጥ አዲስ የብቃት ደረጃ አስመዝግቧል። በልዩ የ"Loop" አርክቴክቸር እና በ128K የቃላት ይዘት (context) የተገነባ ሲሆን፣ ለጥልቅ አስተሳሰብ እና ለትዕዛዝ አፈጻጸም ተብለው የተለዩ "Thinking" እና "Instruct" የተባሉ ዘርፎች አሉት።

@webthreeth
Ethiopia Set to Launch Dedicated AI University

Ethiopia's Prime Minister Abiy Ahmed announced on January 5, 2026, plans for the world's second dedicated AI university after the UAE's MBZUAI, focusing on robotics, machine learning, and data mining, with operations starting in six months. The university will enroll 1,000 students yearly via merit-based entry and provide 100 scholarships to African students.

የኢትዮጵያ ጠቅላይ ሚኒስትር አብይ አህመድ ጃንዋሪ 5 ቀን 2026 በሰጡት መግለጫ፣ ከዓረብ ኤምሬትሱ ኤምቢዜድዩኤአይ (MBZUAI) በመቀጠል የዓለማችን ሁለተኛ የሆነውንና ራሱን የቻለ የአርቴፊሻል ኢንተለጀንስ (AI) ዩኒቨርሲቲ ለመክፈት ዕቅድ መያዙን አበሰሩ። በስድስት ወራት ውስጥ ስራ ይጀምራል ተብሎ የሚጠበቀው ይህ ዩኒቨርሲቲ፤ በሮቦቲክስ፣ ማሽን ለርኒንግ እና ዳታ ማይኒንግ ዘርፎች ላይ ትኩረቱን ያደርጋል። ዩኒቨርሲቲው በየዓመቱ በብቃት ላይ የተመሰረተ መመዘኛን በማለፍ የሚገቡ 1,000 ተማሪዎችን የሚያስተናግድ ሲሆን፣ ለ100 አፍሪካውያን ተማሪዎችም የነጻ ትምህርት ዕድል (Scholarship) ይሰጣል።

@webthreeth
😁2
xAI Raises $20 Billion in Oversubscribed Round Backed by NVIDIA, Cisco

Elon Musk’s xAI secured $20 billion in a Series E funding round on January 6, 2026, surpassing its $15 billion target with strategic investments from technology heavyweights NVIDIA and Cisco. The capital injection is earmarked for expanding the world's largest GPU clusters, a critical infrastructure step needed to train the company’s forthcoming Grok 5 model.

የኢሎን ማስክ xAI ኩባንያ በጃንዋሪ 6 ቀን 2026 በተካሄደው የ"Series E" የገንዘብ ማሰባሰብ ዙር 20 ቢሊዮን ዶላር አሰባሰበ። ይህም ኩባንያው ካቀደው የ15 ቢሊዮን ዶላር ግብ በላይ ሲሆን፣ ከቴክኖሎጂው ዘርፍ ግዙፍ ኩባንያዎች ከሆኑት ኤንቪዲያ (NVIDIA) እና ሲስኮ (Cisco) ስትራቴጂካዊ ኢንቨስትመንቶችን አግኝቷል። ይህ የተገኘው አዲስ ካፒታል በዓለም ላይ ትልቁን የጂፒዩ (GPU) ስብስብ ለማስፋፋት የሚውል ሲሆን፣ ይህም የኩባንያውን ቀጣይ "ግሮክ 5" (Grok 5) ሞዴል ለማሰልጠን ወሳኝ የሆነ የመሰረተ ልማት እርምጃ ነው።

@webthreeth
Ethiopia develops AI and telemedicine to improve Healthcare

The Ministry of Health of Ethiopia has announced a progress in the implementation of artificial intelligence (AI) and telehealth technologies across the country. The revised national health policy also recognises AI and telemedicine as central pillars of future healthcare improvement.

የኢትዮጵያ ጤና ሚኒስቴር በመላ ሀገሪቱ የሰው ሰራሽ አስተውሎት (AI) እና የቴሌጤና (Telehealth) ቴክኖሎጂዎች ትግበራ ላይ መሻሻል መመዝገቡን አስታወቀ። የተሻሻለው ብሔራዊ የጤና ፖሊሲም ሰው ሰራሽ አስተውሎትን እና ቴሌሜዲሲንን (Telemedicine) ለወደፊቱ የጤና አጠባበቅ መሻሻል እንደ ዋና ምሰሶዎች እውቅና ሰጥቷል።

@webthreeth
👎1
OpenAI Launches ChatGPT Health, Integrating Apple Data and Medical Records

OpenAI introduced ChatGPT Health, a dedicated and isolated tab designed to provide personalized medical advice by securely integrating user data from Apple Health and electronic medical records via b.well. The new feature aggregates information from a suite of wellness applications to offer a comprehensive health overview.

ኦፕን ኤአይ (OpenAI)፤ ከአፕል ሄልዝ (Apple Health) እና ከኤሌክትሮኒክ የህክምና መዛግብት የሚገኙ የተጠቃሚ መረጃዎችን በb.well አማካኝነት ደህንነቱ በተጠበቀ መልኩ በማቀናጀት፣ ግላዊ የህክምና ምክር ለመስጠት ታስቦ የተዘጋጀ "ቻትጂፒቲ ሄልዝ" (ChatGPT Health) የተሰኘ ራሱን የቻለ እና የተነጠለ ታብ (Tab) ይፋ አደረገ። ይህ አዲስ አገልግሎት አጠቃላይ የጤና እይታን ለመስጠት ከተለያዩ የጤና አጠባበቅ መተግበሪያዎች የሚገኙ መረጃዎችን ያሰባስባል።

@webthreeth
👎1
DeepSeek To Release Next Flagship AI Model With Strong Coding Ability

DeepSeek's planned V4 AI model was released in early February 2026, with internal benchmarks showing superior coding performance over Anthropic's Claude and OpenAI's GPT series, according to The Information.

እንደ "ዘ ኢንፎርሜሽን" (The Information) ዘገባ፣ የዲፕሲክ (DeepSeek) የታቀደው የ"V4" (ቪ4) AI (ኤአይ) ሞዴል በፌብሩዋሪ 2026 መጀመሪያ ላይ ይለቀቃል። እንደ ዘገባው ከሆነ፣ የውስጥ መለኪያዎች (benchmarks) እንደሚያሳዩት ይህ ሞዴል በአንትሮፒክ (Anthropic) "ክሎድ" (Claude) እና በኦፕንኤአይ (OpenAI) የ"ጂፒቲ" (GPT) ተከታታይ ሞዴሎች ላይ የላቀ የኮዲንግ (coding) ብቃት አሳይቷል።

SWE is cooked, if we have a free version better than Opus 4.5

@webthreeth
👎21👍1
Midjourney Launches Niji V7 with Enhanced Text and Style Features

Midjourney Inc. unveiled Niji V7, its latest AI model for generating anime-style imagery. The upgrade promises sharper coherence, superior prompt adherence, seamless text embedding, and refined style references, enabling users to produce more vivid and consistent visuals.

ሚድጃርኒ ኢንክ (Midjourney Inc.) የአኒሜ ቅጥ (anime-style) ያላቸውን ምስሎች ለመፍጠር የሚያገለግለውን የቅርብ ጊዜ የሰው ሰራሽ አስተውሎት (AI) ሞዴል የሆነውን "ኒጂ ቪ7" (Niji V7) ይፋ አደረገ። ይህ ማሻሻያ የተሻለ ምስልን የማዋቀር አቅም (coherence)፣ ለሚሰጠው የጽሑፍ ትዕዛዝ (prompt) ፍጹም ተገዢነት፣ እንከን የለሽ የጽሑፍ ውህደት እና የተሻሻሉ የስታይል (style) ማጣቀሻዎችን ይዞ መጥቷል። ይህም ተጠቃሚዎች ይበልጥ ደማቅ እና ወጥ የሆነ ጥራት ያላቸውን ምስሎች እንዲፈጥሩ ያስችላቸዋል።

@webthreeth
Anthropic Launches Claude Cowork

Anthropic is launching Claude Cowork (also known as Claude Tasks), a new feature designed to bring autonomous agent capabilities to non-technical users. Available as a research preview for Max plan subscribers on the macOS desktop app, Cowork runs long-duration tasks directly on your computer without requiring terminal commands. Powered by the same technology behind Claude Code, it can access local context and drive web browsers to automate complex workflows seamlessly.

አንትሮፒክ (Anthropic) "ክሎድ ኮወርክ" (Claude Cowork) ወይም "ክሎድ ታስክስ" (Claude Tasks) የተሰኘውን አዲስ አገልግሎት ለመልቀቅ እየተዘጋጀ ነው። ክሎድ ኮወርክ በዴስክቶፕ መተግበሪያ የሚሠራ ሲሆን፣ "ክሎድ ኮድ"ን (Claude Code) በሚያንቀሳቅሰው በዚሁ የሰው ሰራሽ አስተውሎት ወኪል (AI Agent) የሚታገዝ ነው። ይህ አገልግሎት በማክ ኦኤስ (macOS) የክሎድ ዴስክቶፕ መተግበሪያ ለሚጠቀሙ የ"ማክስ" (Max) ፓኬጅ ተመዝጋቢዎች እንደ የጥናት ሙከራ (research preview) ቀርቧል።

@webthreeth
Claude Code Built 'Cowork' Entirely on Its Own in Just 1.5 Weeks

According to Boris Cherny, the entire codebase for Anthropic's new "Cowork" feature was written by Claude Code itself. The AI autonomously generated all the necessary code and had the product ready for shipment in just a week and a half, marking a significant milestone in AI self-development.

ቦሪስ ቼርኒ (Boris Cherny) እንደገለጹት፣ አዲሱ "ክሎድ ኮወርክ" (Claude Cowork) የተገነባው ሙሉ በሙሉ በ"ክሎድ ኮድ" (Claude Code) ነው። ክሎድ ኮድ ይህንን ምርት ለገበያ በሚያበቃ ደረጃ (ready for shipment) ያለ ሰው እገዛ ኮዱን ሙሉ ለሙሉ ጽፎ ለማጠናቀቅ የፈጀበት አንድ ሳምንት ተኩል ብቻ ነው።

@webthreeth
Z.ai Launches GLM-Image and Outperforms Google and OpenAI in Text Rendering Benchmarks

Z.ai has officially released GLM-Image, an open-source model that pioneers "Cognitive Generation" by fusing autoregressive capabilities with diffusion decoders. This architecture allows it to master complex visual tasks like posters and infographics. Crucially, recent performance data reveals that GLM-Image has surpassed industry giants, beating Google’s Nano Banana Pro and OpenAI’s GPT-Image 1 [High]. It secured top scores on both the CVGT-2K and LongText-Bench benchmarks.

ዜድ.ኤአይ (Z.ai) ይፋ ያደረገው "GLM-Image" (ጂኤልኤም-ኢሜጅ) የተሰኘው አዲስ የሰው ሰራሽ አስተውሎት ሞዴል፣ በዘርፉ ግዙፍ ከሆኑት ጎግል እና ኦፕንኤአይ የላቀ ውጤት አስመዝግቧል። ይህ ሞዴል በተለይም "CVGT-2K" እና "LongText-Bench" በተሰኙ የመለኪያ ሙከራዎች (benchmarks) ላይ የጎግልን "Nano Banana Pro" እና የኦፕንኤአይን "GPT-Image 1 [High]" በቃላት ትክክለኛነት እና በረጅም ጽሑፍ አጻጻፍ ብልጫ አሳይቷል።

@webthreeth
👍1
Did you know?

AI-generated cover of Stromae's 2013 hit "Papaoutai". It debuts at #168 on Spotify's Global Daily chart with 1.29 million streams. You can't stop the inevitable. All Ethiopian TikTok edits done this week by an AI song.

https://open.spotify.com/track/5Ue1gLApNgWJmUsQ1z2Ptt?si=8kYmRFVBQWi1-78SZ5cmSQ