Google DeepMind
Google DeepMind is Alphabet's AI division, home of the Gemini model family, the Nobel-recognized AlphaFold and the Veo video generator. Through Google Search, Android and the cloud, its models reach billions of users. This page gathers all our coverage of Google and DeepMind: models, research and products.

Google DeepMind finds softmax flaw: gradient drops to zero at a logit gap of 16.6
Barron, Gnanapragasam and Tschannen show that in float32 the cross-entropy gradient vanishes at a gap of about 16.6, and distillation gets spurious noise.

8B iOS agent beats Gemini-3.5-Flash on UINavBench and matches Gemini-3.1-Pro
An 8-billion-parameter agent trained directly on live iOS apps beat Gemini-3.5-Flash on UINavBench. On AndroidWorld it succeeds almost 7 times as often as its SFT…

A drone with the 1B OpenFlyV2 model flies to its target on its own in 48% of real flights, without training on them
Shanghai AI Lab's OpenFlyV2 packs scene understanding and quadcopter control into a single 1B-parameter on-board model. In real flights, it reaches the target 48%…

Meta releases Muse Spark Open: 816B parameters and the lead among US open-weight models

Switzerland's Apertus 1.5 Gains Vision, Hearing and a 262K-Token Context Window

Goodfire Researcher Claims Neural Networks Are Far Simpler Than Interpretability Experts Thought

BigSleep LLM Agent Finds 45 V8 Vulnerabilities, Fuzzilli 37, With 25% Overlap

Gemma 26B fine-tuned on a 24 GB MacBook: offline assistant solves 703 of 766 conversations
A developer fine-tuned Gemma 4 26B A4B with QLoRA on a laptop, moving a 5,000-token rulebook into the weights. Result: 703 of 766 conversations correct, median 1.64…

Cognition releases FrontierCode 2.0: the best score is just 45.5%
Cognition updated its coding-agent benchmark with larger tasks and stricter criteria. Claude Opus 5.5 scored 45.5%, while Kimi K3 managed only 17.7%, down from…

Google Releases Gemini 4 Argon: $2 per Million Tokens and First Place on the Vals Index

Anthropic Files S-1: $11.6 Billion in Quarterly Revenue and $780 Billion in Commitments

Memory Compaction in AI Agents Wipes Out Up to a Third of Task Rules — and Benchmarks Don't Notice

Retell Grows to 100 Million Calls a Month on OpenAI Models and Codex

Meta opens Muse Spark at 816B, Reflection releases Beam under Apache 2.0
Muse Spark Open scored 39 on Artificial Analysis, making it the strongest American open-weights model. Reflection Beam 1.0 (501B) is second with 30.

DeepSeek V4.1 Flash scores 45 on Artificial Analysis and takes the price lead
Three models shipped October 16–19: DeepSeek V4.1 Flash, MiniMax M3.1 Flash and Mercury 3. We look at what they do, what they cost, and where independent tests…









