OpenAIReasoning, software and computer use.
Official sourceAll research hubs
What can intelligence do?
Models & intelligence
Reasoning, multimodality, model architectures and the evidence behind capability claims.
16 research briefs 96 distinct sources
Research briefs
01
Frontier Reasoning Models & Test-Time Compute
System 2 deliberate reasoning, process reward models, and inference-time search scaling
6 sourcesUpdated 2026-08-18
02Mixture-of-Experts (MoE) & Multi-Head Latent Attention
Sparse activation scaling, auxiliary-loss-free routing, and memory bandwidth optimization
6 sourcesUpdated 2026-08-18
03Context Engineering & Long-Context Architecture
Million-token context windows, needle-in-a-haystack retrieval, prompt caching, and attention sink dynamics
6 sourcesUpdated 2026-08-18
04Reinforcement Learning from Verifiable Rewards (RLVR)
Deterministic reward oracles, formal theorem verification, and self-directed policy optimization
6 sourcesUpdated 2026-08-18
05Multimodal Reasoning & Vision-Language-Action Models
Native multimodality, spatial intelligence, 4D world representations, and cross-modal attention
6 sourcesUpdated 2026-08-18
06Post-Training Distillation & Speculative Decoding
Knowledge transfer from reasoning teachers, speculative token drafting, and edge model optimization
6 sourcesUpdated 2026-08-18
07Synthetic Data Curation & Automated Curricula
Constitutional AI, self-instruct generation, automated filtering, and epistemic quality metrics
6 sourcesUpdated 2026-08-18
08Diffusion Transformers & Generative Neural Video
DiT architectures, flow matching, spatio-temporal attention, and cinematic video synthesis
6 sourcesUpdated 2026-08-18
09Neural Audio Synthesis & Conversational Speech Models
Neural audio codecs, zero-shot voice cloning, expressive prosody modeling, and full-duplex conversational audio
6 sourcesUpdated 2026-08-18
10Sparse Attention, State Space Models & Linear Transformers
Mamba 2, RWKV, FlashAttention-3, sub-quadratic attention, and hybrid state-space transformer backbones
6 sourcesUpdated 2026-08-18
11Representation Engineering & Mechanistic Interpretability
Sparse Autoencoders (SAEs), concept steering vectors, dictionary learning, and circuit analysis
6 sourcesUpdated 2026-08-18
12Adversarial Robustness & Jailbreak Defense Architectures
Automated red-teaming, prompt injection defenses, multi-turn jailbreak mitigation, and robust alignment
6 sourcesUpdated 2026-08-18
13Multilingual Frontier Intelligence & Cross-Lingual Transfer
Polyglot tokenizers, cross-lingual representation alignment, low-resource transfer, and cultural nuance
6 sourcesUpdated 2026-08-18
14Mathematical Theorem Proving & Formal Verification AI
Lean 4, Isabelle provers, AlphaProof, AlphaGeometry, and neuro-symbolic automated deduction
6 sourcesUpdated 2026-08-18
15Embodied Physical AI & Spatial World Models
Humanoid robotics policies, spatial simulation, physics world representations, and end-to-end tactile learning
6 sourcesUpdated 2026-08-18
16Enterprise AI Model Strategy: Build, Fine-Tune, or Buy?
Capital-aware decision frameworks for API rental, retrieval, adapter fine-tuning, and sovereign hosting
6 sourcesUpdated 2026-08-18
Inside Models & intelligence
The labs. The current models.
Selected model families, checked against provider documentation on 13 September 2026. Specifications are provider-reported; local experiments remain separately dated.
OpenAI
AnthropicLong-horizon agents, coding and reasoning.
Official source
GoogleMultimodal systems and agentic workflows.
Official sourcexAIxAI
Configurable reasoning and tool use.
Official source