The AI Game NPC Revolution: LLM-Driven NPCs Transform the Player Experience
LLM-driven game NPCs enable free-form dialogue and dynamic storylines; several AAA titles have already adopted them.
Google Releases Gemini 3, Embedding It in Search on Day One
Google released Gemini 3 on November 18, 2025, calling it its most intelligent model — and for the first time embedded it in Search on launch day.
Qwen3-Max Released: Alibaba's First Trillion-Parameter Model, Pretrained on 36T Tokens
Alibaba released Qwen3-Max on September 24, 2025 — its largest model ever at over 1 trillion parameters — alongside Nvidia partnership news.
OpenAI Officially Releases GPT-5 with Built-in Thinking, Free for All Users
OpenAI released GPT-5 on August 7, 2025, calling it its 'smartest, fastest, most useful model yet,' with built-in thinking available to free users.
Moonshot AI Releases Kimi K2: A Trillion-Parameter MoE Model, Open-Sourced
Moonshot AI unveils Kimi K2 with 1 trillion total parameters — the largest open model at launch — excelling at agentic and coding tasks.
Qwen 3 Released: MoE Architecture Sets a New Bar for Open Source
Alibaba releases the Qwen 3 series; the flagship Qwen3-235B-A22B rivals top models like DeepSeek-R1 and o1 on major benchmarks.
Meta Releases Llama 4: Scout and Maverick Debut as Natively Multimodal Open Models
Meta released Llama 4 Scout and Maverick on April 5, 2025 — its first natively multimodal open-weight models, with Scout supporting a 10M-token context.
Google Releases Gemini 2.5 Pro: Its First Thinking Flagship Tops the Benchmarks
Google released Gemini 2.5 Pro in March 2025 with built-in reasoning, leading multiple benchmarks by meaningful margins.
DeepSeek V3 Released: A 671B MoE Model That Shook the Industry on Cost
DeepSeek released V3 on December 26, 2024: 671B total / 37B active parameters, trained for a fraction of closed-model budgets, open and commercial-ready.