General-purpose foundation models
Large models that reason, code, converse and use tools across most cognitive domains at or above expert-human level on many benchmarks.
State of the art
Frontier LLM agents (2025–2026) perform multi-hour software and research tasks with tool use; reliability over days-long tasks is unproven.
- Reliability and calibration over long horizons
- Continual learning after deployment
Who’s working on it
- OpenAI ↗Startup
GPT-5-series models (2025-2026) lead many agentic coding and reasoning benchmarks
- Anthropic ↗Startup
Claude Opus models (2025-2026) recorded some of the longest METR task time horizons
- Google DeepMind ↗Company
Gemini 3 (Dec 2025) and Gemini 3.1 Pro (2026) are frontier multimodal reasoning models
- DeepSeekStartup
Chinese lab whose open-weight R1 reasoning model (Jan 2025) matched frontier results at reported low training cost
- Demis HassabisResearcherGoogle DeepMind
Leads Google DeepMind's Gemini program; 2024 Nobel Prize in Chemistry for AlphaFold
- METR ↗Academic lab
Independent evaluator measuring how long the tasks are that frontier agents can complete (Time Horizon 1.1, Jan 2026)
What to invest in
| Kind | Investment | Scale |
|---|---|---|
| research | Long-horizon reliability, memory and continual learning | $$$$ > $1B |