Geo News
Community curated by people like you
LatestAICryptoHealthWorld AffairsUS Politics

Mixture of experts (MoE) stories

Oct 5, 2026

Reflection AI unveils Beam open-weight model to compete with Chinese AI

Nvidia-backed Reflection AI debuted Beam, a 501B parameter sparse MoE model with 23B active parameters, designed to compete with lower-cost Chinese models like DeepSeek in coding and agentic tasks.

Oct 5, 2026·5 sources
00
Oct 4, 2026

Aleph Alpha releases Kolibri 78B open-weight German-English model

Aleph Alpha released Kolibri, a 78.1B parameter mixture-of-experts model for German and English with only 3.46B active parameters, as an open-weight release.

Oct 4, 2026·2 sources
00
Oct 2, 2026

Ai2 releases OlmoCore 3 training framework for mixture-of-experts models

Allen Institute for AI released OlmoCore 3, an open development framework for large mixture-of-experts language models that delivers 2.7x throughput gains.

Oct 2, 2026·2 sources
00
Aug 28, 2026

Tencent releases Hy4 preview, open-source AI model with self-improvement capabilities

Tencent has unveiled Hy4 Preview, a 770-billion-parameter open-source Mixture-of-Experts model that uses a recursive self-improvement loop to boost inference throughput by 31.8%. The release, trained on Tencent's vast product ecosystem, positions the company back in the top tier of China's AI competition.

Aug 28, 2026·3 sources
00
Aug 26, 2026

Zhipu AI reveals GLM-5.3-Flash model ran on 100,000 Chinese chips

China's Zhipu AI confirmed that its viral Ox Alpha model, now revealed as GLM-5.3-Flash, is a 320B-parameter mixture-of-experts AI system that ran entirely on 100,000 domestically produced Chinese chips during launch. The geopolitically significant claim about Chinese chip capabilities remains unverified by independent sources.

Aug 26, 2026·4 sources
00
Jul 30, 2026

Thinking Machines Releases Inkling-Small, a 276B Parameter Mixture-of-Experts Model

Thinking Machines has released Inkling-Small, an open-weights Mixture-of-Experts transformer model with 276 billion parameters that achieves performance comparable to its larger predecessor Inkling while being one-quarter the size.

Jul 30, 2026·1 source
00
Jul 15, 2026

Thinking Machines Lab Releases Inkling, 975B-Parameter Open-Weights Multimodal AI Model

AI startup Thinking Machines Lab, led by former OpenAI CTO Mira Murati, has released Inkling, a 975-billion parameter mixture-of-experts model with open weights designed for enterprise customization and on-premises deployment. The company explicitly positions the model as focused on customizability rather than benchmark dominance.

Jul 15, 2026·7 sources
00
Jun 22, 2026

Tokyo's Sakana AI launches Fugu Ultra, multi-model orchestration system matching frontier AI performance

Sakana AI unveiled Fugu and Fugu Ultra on June 22, 2026, a multi-agent orchestration system that coordinates multiple AI models to match performance of leading models like Anthropic's Fable 5 and Mythos without training its own frontier model. The approach aims to reduce single-vendor dependency in enterprise AI deployments.

Jun 22, 2026·8 sources
00
May 7, 2026

Zyphra Releases ZAYA1-8B Reasoning Model Trained on AMD Hardware

Zyphra launched ZAYA1-8B, a Mixture of Experts reasoning model with only 760M active parameters that reportedly outperforms much larger open-weight models on math and coding benchmarks, trained entirely on AMD hardware.

May 7, 2026·1 source
00

Top claims

  • ▪Beam was pretrained on 23.8 trillion tokens from web, public, and proprietary licensed datasets, with pretraining completed in under four weeks on 6,144 NVIDIA GB300 GPUs
  • ▪NVIDIA CEO Jensen Huang has defended the use of Chinese open-weight models by US businesses, stating they provide good value for companies
  • ▪Highly efficient, smaller AI models can successfully compete with larger frontier models

People involved

Mira Murati

Subtopics

Open-source AI7AI foundation models5AI startups3Compute, chips & AI infrastructure3Open weights vs closed models3AGI arms race & geopolitics2AI agents2Business & enterprise AI2China tech & industrial strategy2Large language models (LLMs)2AI1AI coding assistants1AI inference scaling1AI research & benchmarks1Allen Institute for AI1China AI regulations1Multilingual models1Multimodal models1Open model deployment tools1Reasoning models1Recursive self-improvement1

Related timelines

Congress

110 stories

Crypto hacks

100 stories

Iran War

209 stories

Payments

118 stories

Russia-Ukraine war

136 stories