Frontier systems combined diagrams, PDF papers, and code in one tool-verified reasoning thread.
In early 2026, demos showed agents reading papers with figures, writing experiment code, and formalizing intermediate lemmas.
Integration is imperfect: scientific OCR and diagram errors persist. Still, assisted-researcher throughput grew.
The pattern feeds both product labs and AI-for-pure-math efforts.
OpenAI's new multimodal models integrate text, image, and audio with refreshed APIs for developers.
Announced ~10 September 2026, DeepSeek-V4.1-Flash is a native multimodal (vision+text) 552B MoE with an asymmetric causal encoder–decoder (8B active on input, 16B on output), sharply smaller KV cache, peak/off-peak pricing, and retirement of V4-Flash—direct pressure on closed labs’ cost per token.
Models optimized to think longer at inference defined 2025, with leaps in math contests and coding agents.