AI benchmarks — Page 2

33 stories tagged AI benchmarks.

A robot sits at a workbench with scientific equipment and an atom symbol, illustrated in blue and orange tones with "WHY WE…

AI Benchmarks Are Breaking. Here's Why That Matters.

New ARC-AGI-3 benchmark exposes how AI models memorize rather than learn. Humans score 100%, frontier AI models score less than 1%. The gap reveals everything.

A presenter on stage introduces GPT 5.4 Pro, with a futuristic white and green robot head displayed on the left and glowing…

GPT-5.4 Pro Costs $180 Per Million Tokens—And Beats Google at Its Game

Bob Reynolds6 months ago

AI. Bob Reynolds6 months ago

GPT-5.4 Pro Costs $180 Per Million Tokens—And Beats Google at Its Game

OpenAI's GPT-5.4 Pro outperforms competitors on new benchmarks, but at a steep price. What the latest AI model tells us about the real race.

Retro-styled control room with three humanoid robots monitoring data charts and screens, displaying exponential growth…

AI Agents Are Accelerating—But Nobody Agrees What That Means

Dev Kapoor7 months ago

AI. Dev Kapoor7 months ago

AI Agents Are Accelerating—But Nobody Agrees What That Means

New benchmarks show AI coding agents tripling capabilities in months. Researchers urge caution. Investors price in economic collapse. Welcome to 2026.

Three stylized robots with Google logos hold various tools against a colorful gradient background with sparkles and…

Google's Gemini 3.1 Pro: When Benchmark Wins Stop Mattering

Bob Reynolds7 months ago

AI. Bob Reynolds7 months ago

Google's Gemini 3.1 Pro: When Benchmark Wins Stop Mattering

Gemini 3.1 Pro tops AI benchmarks, but the real story is cost efficiency and multimodal capabilities—not another 'world's most powerful model' claim.

Man in business suit speaking at microphone in OpenAI office with yellow text overlay reading "ONLY 2 YEARS LEFT...

Sam Altman Says AGI Arrives in 2 Years. Here's the Data.

Tyler Nakamura7 months ago

AI. Tyler Nakamura7 months ago

Sam Altman Says AGI Arrives in 2 Years. Here's the Data.

OpenAI's Sam Altman just compressed the AGI timeline to 2028. We examined the benchmarks, the skepticism, and what 'world not prepared' actually means.

Man with surprised expression next to AI model selection interface featuring Gemini 3.1 Pro marked as "NEW" with pricing…

Google's Gemini 3.1 Pro: Genius on Paper, Disaster in Practice

Zara Chen7 months ago

AI. Zara Chen7 months ago

Google's Gemini 3.1 Pro: Genius on Paper, Disaster in Practice

Gemini 3.1 Pro crushes benchmarks but fails at basic tasks. Developer Theo tests Google's 'smartest model ever' and finds a genius that can't follow instructions.

Man with glasses in black blazer against white background with text "THE DOWNFALL OF AI BENCHMARKS

Why AI Benchmarks Are Breaking (And What That Means for You)

Tyler Nakamura7 months ago

AI. Tyler Nakamura7 months ago

Why AI Benchmarks Are Breaking (And What That Means for You)

Google's Gemini 3.1 Pro drops alongside a bigger question: are AI benchmarks even measuring what we think they are? The answer affects your buying decisions.

Google's Gemini 3.1 Pro: Testing the Hype vs. Reality

Rachel "Rach" Kovacs7 months ago

AI. Rachel "Rach" Kovacs7 months ago

Google's Gemini 3.1 Pro: Testing the Hype vs. Reality

Google's Gemini 3.1 Pro shows impressive benchmark gains and coding abilities, but real-world testing reveals persistent issues that temper the enthusiasm.

Futuristic AI presentation featuring a cyborg figure with neon green slime dripping from its head, white headphones, and…

Chinese AI Models Are Suddenly Catching Up—And Fast

Zara Chen7 months ago

AI. Zara Chen7 months ago

Chinese AI Models Are Suddenly Catching Up—And Fast

GLM-5 claims to beat major US models on reliability while open-source agents hit near-human scores. The AI race just got a lot more complicated.

Man in dark shirt next to text reading "VIBE CODING WORKING" with striped red text and AI logos on dark blue background

Claude Opus 4.6 Is Smarter—And Vastly More Expensive

Rachel "Rach" Kovacs7 months ago

AI. Rachel "Rach" Kovacs7 months ago

Claude Opus 4.6 Is Smarter—And Vastly More Expensive

Anthropic's newest AI model excels at knowledge work but burns through tokens 60% faster than its predecessor—and passed a benchmark by lying and forming cartels.