Google DeepMind

Company-reported MTEB Code rose from 68.76 to 78.68; context is 8K (4 times its predecessor), and MRL lets the 768 dimensions shrink to 128, cutting storage by up to 6 times. Quantized, it runs on a Pixel 11 Pro in about

Company-reported scores: DeepSWE v1.1 77.9%, AutomationBench 51.3% (first), LVBench 91.7%, CWE-bench v1 68% (tied first). The introductory price is $2/$10 per 1 million tokens (later $4/$20), and Google said it is taking

A case of building a public good to prepare for the next outbreak by loading AI-generated predicted structures wholesale into a public database. The starting point of pandemic response shifts from "solving structures thr

Sells voice not as "transcribe, then process the text" but split into two tiers of model with adjustable thinking budgets — a form that separates latency and reasoning depth into products within the same interface.

An extension of the AlphaFold DB pattern of precomputing and publishing a model's outputs in full, rather than offering access to the model. The company claims early collaborators found 22% more non-coding genetic associ

AI weather models move beyond research demos to become the default forecasts in consumer products and commercial APIs

Security-specific derivative models are split off as official variants of the frontier lineup — vulnerability detection and patching become an independent axis of the model family

An attempt at evaluation infrastructure that aims both to prevent benchmark contamination and to protect IP — a path to checking frontier models with private benchmarks

Shows a path to obtaining a diffusion model with less than 10% of the original token budget; about 1,500 tok/s on a single H100

The core VLA model is early-access only, while ER 2 is released through AI Studio

Real progress in domains that can be checked automatically (Google's own measurement)

ARC-AGI-2 is effectively saturated. The announcement does not discuss AGI, framing the upgrade instead as accelerating science, research and engineering

The ARC-AGI-2 SOTA doubled in a single leap

Establishes a two-track release strategy: only ER is opened via API, while the VLA model itself stays closed

Adds harmful manipulation as a new CCL and tackles misalignment, such as resistance to shutdown, head-on

A rebuttal showing that the limits of single-vector retrieval are fundamental, dimensional limits that more data or bigger models cannot fix. A theoretical basis for a return to multi-vector, sparse and hybrid retrieval

Generates controllable environments from a prompt, staying consistent for several minutes at 720p (720 lines) and 24 fps

That a general-purpose model achieved this end-to-end in natural language, without formalization, is the key piece of evidence in the debate over generality

A consensus document on CoT monitoring, co-signed by major labs

A position paper that makes the practice of using CoT as an explanation its subject

The start of the lineage of automated algorithm discovery

Connected web-scale semantic knowledge to robot control as an output modality

Represents the shift of embedding models from small dedicated encoders to derivatives of frontier LLMs

The de facto default vision encoder for VLMs. Multimodal embeddings and VLM backbones converged into the same lineage

Google claimed 2.0 Flash outperforms 1.5 Pro on key benchmarks at twice the speed, launched the Multimodal Live API and Deep Research, and presented agent prototypes such as Project Astra, Project Mariner and Jules

The first case of provenance marking for generated content moving beyond research into large-scale deployment

Self-training grounded in a formal language reached olympiad level. DeepMind stated explicitly that it would not have been possible without Lean and Mathlib

Framed the long-context vs RAG debate as a cost-performance trade-off

Together with RSP and PF, established a system of three major risk-evaluation frameworks

Extended beyond proteins to complexes with nucleic acids, small molecules and ions

The starting point of the world-model lineage. The authors argue that at this scale it can be regarded as a foundation world model

Context length jumped by an order of magnitude, sparking the RAG-is-dead debate

Suggested that CoT may be a decoding problem rather than a prompting one

Handed the very choice of prompting technique over to the model

An early case of formally verifiable rewards driving reasoning training. Produces human-readable proofs

A symbol of AI-driven materials discovery, and later the subject of the biggest reproducibility dispute

The starting point of shared cross-embodiment datasets

A branch that inserts an abstraction step through the prompt

A direct rebuttal of the Self-Refine/Reflexion family

A branch that gets the few-shot effect without labeling examples

The evolutionary prompt-optimization lineage that leads to GEPA

Formalized automatic prompt optimization as a general optimization problem

Transferred web-scale VLM knowledge directly to robot control, establishing the very category of VLA

Redefined linear CoT as a search problem

The de facto successor to CLIP in multimodal embeddings. Removing the need for very large batches greatly lowered the barrier to training

An independent formulation of the same technique, with a 2–2.5× speedup

A lineage of modular, decomposed reasoning rather than single CoT generation

The prototype of the few-shot VLM, connecting frozen vision and language models through cross-attention

Exposed that the large models of the time were undertrained; the 70B model outperformed the 280B Gopher

An alternative that processed sequences with dilated convolutions instead of recurrence. One branch of the race to move beyond recurrence

Showed that autoregressive generation extends to audio. The foundation for later TTS and audio generation

An event that changed public perception of what AI can do. It became the backdrop for later discussions of investment and policy

Revealed the combination of policy and value networks with Monte Carlo tree search. Part of the lineage that combines search and learning

The point at which deep reinforcement learning passed peer review and was recognized as an official achievement

The idea of handling memory through attention-based addressing. The start of the memory-augmented lineage

An early case showing that deep reinforcement learning works from screen input alone