AI Reasoning Models Hit Performance Wall Despite Higher
New research reveals AI reasoning models are hitting performance walls on creative tasks while dramatically increasing…
Breaking news, research, and analysis on artificial intelligence — covering the models, methods, and shifts shaping the field.
New research reveals AI reasoning models are hitting performance walls on creative tasks while dramatically increasing…
Miami startup Subquadratic claims 1,000x efficiency gains through subquadratic architecture, while Google releases speculative decoding for…
Zyphra released ZAYA1-8B, an efficient 8B-parameter model trained on AMD GPUs, while Subquadratic claims 1,000x efficiency…
DeepSeek released its V4 model achieving near state-of-the-art AI benchmark performance while pricing API access at…
Poolside launched free Laguna XS.2 for local coding agents, SenseTime released U1 image model with direct…
Miami startup Subquadratic claims 1,000x AI efficiency gains through subquadratic architecture while Google and NVIDIA release…
AI reasoning models are driving up compute costs through test-time processing while revealing persistent gaps in…
xAI launched Grok 4.3 with built-in chain-of-thought reasoning at $1.25 per million input tokens, targeting enterprise…
XPENG launched its VLA 2.0 autonomous driving system as a production-ready product in March 2026, while…
Poolside launched free Laguna XS.2 models for coding tasks while Xiaomi released efficient MiMo-V2.5 variants, intensifying…
Poolside launched open source Laguna XS.2 for coding as data shows image AI models now drive…
DeepSeek released V4, a 1.6-trillion-parameter AI model that matches frontier performance at one-sixth the cost of…