AI by field
What AI can do in your field, right now.
Each field's newest models move faster than anyone can follow. Here is where they stand in each one, how fast that is changing, and every milestone marked by how firm it is: shown, claimed, reported or done by a model that is not public. Written fresh every day.
Open your field Free account. No card.
- AI for mathMoving: very fastAI models now solve research-level math problems and formalize proofs, with Claude Fable 5.2 leading benchmarks.ClaimedAnthropic says Claude helped solve two open physics problems and formalize proofs with 1,000 agents.Best here now: Claude Fable 5.2
- AI for biology and drug discoveryNo benchmark of its own yetAI models now design proteins and predict variant effects, but no AI-designed drug is FDA-approved.ClaimedGoogle DeepMind announced AlphaGenome 2, a DNA sequence model predicting variant effects in specific cell types.
- AI for medicine and healthMoving: fastAI models now match or beat physicians on some diagnostic benchmarks, but real-world clinical reliability and regulatory approval remain limited.ShownMicrosoft's MAI-DxO system outperformed physicians on complex diagnostic cases, achieving more than twice the accuracy.Best here now: Claude Sonnet 5.5
- AI for physics, chemistry and materialsMoving: very fastAI models now solve graduate-level science problems and assist in real experiments, but still struggle with deep chemistry reasoning.ClaimedAn unreleased Anthropic model proved the 20-year-old Aaronson–Ambainis quantum conjecture, verified by Lean.Best here now: Claude Fable 5.2
- AI for electrical engineering and chipsNo benchmark of its own yetAI models now assist chip and PCB design, with Claude Fable 5.2 shown to design working electronic prototypes end-to-end.ShownClaude Fable 5.2 achieved end-to-end electronic prototype design, producing functional devices without circuit corrections.
- AI for software engineeringMoving: fastClaude Fable 5.2 leads software engineering benchmarks, with models now solving 93% of SWE-Bench Pro tasks.ReportedGoogle DeepMind released Gemini 4 Carbon for general usage with gains in software engineering.Best here now: Claude Fable 5.2
- AI for game developmentNo benchmark of its own yetClaude Fable 5.2 can generate playable game prototypes from short prompts, but full production remains out of reach.ShownMeshy Rigging on fal turns humanoid 3D models into characters ready to animate, works with Unreal Engine, Unity, Blender, and Mixamo, and prepares fingers for animation.
- AI for content and media productionNo benchmark of its own yetAI now generates high-resolution images and editable 3D scenes from text, with MiniMax H3-Image and GPT Image 2.6 leading.ShownIntangible's Canvas turned written briefs into editable 3D scenes for AI video generation.
- AI for financeMoving: fastAI models now handle real financial analysis tasks, with Claude Fable 5.2 leading finance benchmarks.ShownNorthern Trust and Microsoft launched an AI agent that converts historical client transactions into standardized reporting data, cutting preparation time by 83%.Best here now: Claude Fable 5.2
- AI for lawMoving: fastAI for law is advancing rapidly, with Claude Sonnet 5.5 leading law benchmarks and Gemini 4 Argon topping agentic legal work, but real-world contract review remains below 60%.ShownQwen3.8 Max leads the Access to Justice benchmark with 76.37% rubric criteria met.Best here now: Claude Sonnet 5.5
- AI for remote and knowledge workMoving: fastClaude Fable 5.2 now completes 41.3% of real freelance projects on the Remote Labor Index, up from 2.1% a year ago.ShownOpen-source NorthCUA computer-use models are released, scoring up to 88 on OS World Verified.Best here now: Claude Fable 5.2
- AI for cybersecurityMoving: fastAI models now find and exploit real vulnerabilities at scale, with Claude Mythos 5.2 and GPT-6 Cyber leading.ShownBypassGym benchmark shows even lightweight models like DeepSeek-V4.1 Flash can generate successful EDR evasion submissions.Best here now: GPT-6 Astra
- AI for robotics and the physical worldNo benchmark of its own yetAI models now control real robots and dexterous hands, with NVIDIA's GR00T 2 and Google's Gemini Robotics 2.5 leading in manipulation and industrial tasks.ShownRunway releases Praxis-1, a world action model that ranks first on NVIDIA's RoboLab-120 leaderboard, with open weights under Apache 2.0.
- AI for research and analysisMoving: very fastAI research agents now score 73-92% on hard research benchmarks, led by Claude Fable 5.2 and Kimi K3.1 Pro.ShownLongCat-2.5 leads the ResearchRubrics benchmark at 81.0%.Best here now: Claude Opus 5