Recent posts

Usman Salis Usman Salis Benchmarks
4 hours ago

OpenAI GPT-5.4 Nano Leads AI Cost-Performance Race

OpenAI introduced GPT-5.4 mini and nano, offering lower-cost AI models with strong reasoning capabilities. Benchmarks show nano emerging as a leader in cost-efficiency despite trade-offs.

Peter Smith Peter Smith Industry
5 hours ago

ByteDance Cuts LLM Hallucinations With New RL Method

ByteDance Seed researchers introduced behavioral calibration, a reinforcement learning method that reduces hallucinations in LLMs. The approach enables models to recognize uncertainty and avoid incorrect answers

Saad Ullah Saad Ullah Industry
6 hours ago

T-MAP Uncovers Multi-Step Vulnerabilities Across LLM Agent Systems

A new study introduces T-MAP, a trajectory-aware method exposing critical security flaws in LLM agents operating through multi-step tool interactions in real-world AI environments.

Victoria Bazir Victoria Bazir Industry
9 hours ago

AI Breakthrough: SpecEyes Boosts LLM Speed by Up to 3.35x

A new speculative perception framework for agentic multimodal AI delivers speed gains of up to 3.35x while pushing benchmark accuracy from 81% to 84%.

Eseandre Mordi Eseandre Mordi Benchmarks
9 hours ago

AI Agents Beat cuDNN by 3.5% in NVIDIA's New AVO Framework

NVIDIA unveils Agentic Variation Operators (AVO), showing measurable performance gains over cuDNN and FlashAttention-4 on Blackwell GPUs.

Eseandre Mordi Eseandre Mordi Industry
10 hours ago

ARC-AGI-3 Scoring Overhaul: New Squared Metric Makes Old Benchmarks Incomparable

A new explanation of ARC-AGI-3 reveals a shift toward efficiency-based scoring using a squared metric, making results from earlier ARC-AGI benchmarks not directly comparable.

Alex Dudov Alex Dudov Markets
1 day ago

GPT Models Drop Up to 9% in AI Trading Test While Opus 4.5 Gains 9.4%

AI trading test pits top models vs S&P 500. GPT lags as Opus 4.5 gains 9.4% and GPT 5.1 drops 9%.

Alex Dudov Alex Dudov AI models
1 day ago

Meta's HyperAgents: Recursive Self-Improvement Across 4 Computable Domains

Meta's HyperAgents framework enables recursive AI self-improvement, modifying code, logic, and architecture autonomously.

Marina Lyubimova Marina Lyubimova Industry
1 day ago

HyperAgents AI Learns to Improve Itself Across 3+ Domains

A new AI framework called HyperAgents lets systems improve not just what they do, but how they learn — delivering compounding performance gains across multiple tasks and domains.

Usman Salis Usman Salis Trends
1 day ago

Claude AI Usage Shifts: Experienced Users Drive 48.9% Work Adoption

Anthropic data shows experienced Claude users shift toward work tasks, with personal queries rising and coursework use falling.

Peter Smith Peter Smith Industry
1 day ago

DeepImageSearch: Dual-Memory AI Retrieves Images Across Visual Histories

DeepImageSearch uses dual-memory AI to retrieve images by context, not just content, outperforming current models.

Usman Salis Usman Salis Industry
1 day ago

Tencent's LongHorizonUI Can Now Handle Tasks With 15+ Steps

Tencent and its academic partners have built LongHorizonUI, an AI framework targeting a core weakness in today's agents: falling apart mid-task when workflows get long. Here's how it works.

Saad Ullah Saad Ullah AI models
1 day ago

Grok 4.20 Posts 22% Hallucination Rate, Lowest Among Leading AI Models

Grok 4.20 records the lowest hallucination rate at 22%, beating Claude 4.5 Haiku, MiniMax V2 Pro, and GLM-5 in factual accuracy.

Eseandre Mordi Eseandre Mordi Benchmarks
1 day ago

OpenResearcher AI Scores 54.8% on BrowseComp-Plus, Beating GPT-4.1 and Claude Models

OpenResearcher scores 54.8% on BrowseComp-Plus, beating GPT-4.1 and Claude models with fully offline operation and no reliance on external APIs.

Victoria Bazir Victoria Bazir Markets
1 day ago

Nvidia, Microsoft Drive AI Push Into Nuclear Energy Sector

Microsoft and Nvidia deploy AI tools to cut nuclear permitting time, slash costs, and modernize plant operations.

Eseandre Mordi Eseandre Mordi Industry
1 day ago

AI Breakthrough: TextPecker Boosts Image Text Accuracy by 8.7%

TextPecker framework improves AI text rendering with 8.7% semantic alignment gain and 4% structural fidelity boost.

Saad Ullah Saad Ullah Industry
1 day ago

Tesla and SpaceX Plan a 1-Terawatt Terafab AI Chip Factory in Texas

Tesla, SpaceX, and xAI unveil Terafab, a Texas chip factory targeting 1TW of AI compute capacity.

Alex Dudov Alex Dudov AI models
1 day ago

Zhipu AI Launches ZClawBench With 116 Test Cases for Real-World AI Agent Evaluation

Zhipu AI's ZClawBench tests 116 AI agent cases across coding, automation, and data analysis.

Marina Lyubimova Marina Lyubimova Industry
1 day ago

AI Data Security Tools Expand With Lyo's 5-Step Defense Platform

Lyo launches AI-native platform with 5-step lifecycle to monitor, govern, and remediate data access across enterprise AI environments.

Usman Salis Usman Salis Industry
1 day ago

Alibaba Launches Omni-WorldBench to Test 4D AI in Real-World Scenarios

Alibaba's Omni-WorldBench tests 4D AI in dynamic, real-world spatial and temporal scenarios.

Posts feed
OpenAI GPT-5.4 Nano Leads AI Cost-Performance Race
4 hours ago
ByteDance Cuts LLM Hallucinations With New RL Method
5 hours ago
T-MAP Uncovers Multi-Step Vulnerabilities Across LLM Agent Systems
6 hours ago
AI Breakthrough: SpecEyes Boosts LLM Speed by Up to 3.35x
9 hours ago
AI Agents Beat cuDNN by 3.5% in NVIDIA's New AVO Framework
9 hours ago
ARC-AGI-3 Scoring Overhaul: New Squared Metric Makes Old Benchmarks Incomparable
10 hours ago
GPT Models Drop Up to 9% in AI Trading Test While Opus 4.5 Gains 9.4%
1 day ago
Meta's HyperAgents: Recursive Self-Improvement Across 4 Computable Domains
1 day ago
HyperAgents AI Learns to Improve Itself Across 3+ Domains
1 day ago
Claude AI Usage Shifts: Experienced Users Drive 48.9% Work Adoption
1 day ago
Follow us on Twitter

Disclaimer: Aigazine provides content for informational and educational purposes only. We do not offer financial, investment, or legal advice. Any decisions made based on the information published on this website are solely at the reader’s own discretion and risk. We encourage all readers to conduct their own research and seek professional guidance when necessary.

Aigazine is a news platform focused on artificial intelligence, covering global AI trends, technology, and innovation. Aigazine is based in Tbilisi (0179, Georgia, Tbilisi City, Vake District, 49 Besarion Zhghenti Street, VAT 305786600).

© 2025 aigazine.com

We use cookies to improve your experience on our site and to show you relevant advertising. To find our more, read our privacy policy and cookie policy