Odyssey launches Agora-2, a multi-agent world model supporting 20 humans and agents
rohanpaul_ai · x · 2026-09-25
Odyssey (CEO Oliver Cameron) released a playable research preview of Agora-2, its next-gen multi-agent world model. Key points:
- 5x capacity: up to 20 humans and AI agents inside one shared, real-time simulated environment (Agora-1 supported 4)
- Multiple environments instead of a single one, with richer interactions and longer-horizon behaviors
- Trained on Diablo II: observation-action-state captures let the model learn how entities navigate and respond to each other — effectively a learned game engine
- Target applications: AI training, robotics, autonomous vehicles, defense, energy, cybersecurity and gaming; the team cites AI-powered cyberattacks and agent collusion as reasons studying multi-agent interaction is urgent
A playable preview and technical report are available now.
More from Models
- Predicting a new GPT-4 / Sonnet 3.5 / Opus 4.5 moment from upcoming GPT and Claude releases — kieranklaassen · 2026-09-25
- Anthropic engineer: crank Claude's effort setting for risky code, dial it down for routine tasks — every · 2026-09-25
- Proteus, accepted at NeurIPS 2026, unlocks memory blocks to boost long-context by +8.4 NIAH — behrouz_ali · 2026-09-25
- LangChain launches LangSmith Fine-Tuning with smithtune CLI to train models from agent traces — LangChain · 2026-09-25
- xAI's "xhigh latest" on Pro tier panned as merely "Qwen-tier" — teortaxesTex · 2026-09-25
- AIRA₂ Research Agents Hit 81.5% on MLE-bench-30, Beating Prior SoTA of 72.7% — mariofilhoml · 2026-09-25