Search for a command to run...
Releases that pushed capabilities forward and hints at what comes next
GPT-3 series model, capable of writing coherent, intelligent text. Exceptional at marketing content with few-shot examples demonstrating format and style. Fine-tuning creates apps ecosystem.
Instruction-tuned model demonstrating modern post-training foundations (SFT, RLHF) for tasks rather than completions, and provides a glimpse at future model interactions.
A marked leap in abilities and intelligence – MoE, vision, and expensive 32k context window – accelerating the frontier model wave. Becomes the comparative reference model for years.
Where Claude 1 was largely forgettable, Claude 2 and Claude 2.1 showcase Anthropic's unique approach to humanistic language generation and a massive 200k token context window.
Builds upon GPT-4's weaknesses, increasing intelligence and reimagines the model with 128k context length, function calling for tool usage, vision, and high TPS – in one model.
xAI trains the first frontier model on one co-located supercomputer cluster with 200k GPUs, demonstrating that – despite the fear and rumors – pre-training scaling laws are far from over.
OpenAI's second generation reasoning model sets the bar for agentic behavior and intelligence. Classic benchmarks are neck-and-neck, but real-world use shows 1-2 OOM intelligence.
Gemini Flash 2.5 offers lightning-fast responses, 1M context, and near-zero cost intelligence. Google perfectly rides the price:performance Pareto frontier providing developers better app margins.
Future milestones expected with 145-155 IQ at 1:10 the cost, continual model learning, human-level audio and vision capabilities, and exceptional intelligence capable of novel discovery.