Trendingby GlyphMixture of experts
Peak · emerged January 2017 · high July 2026 · +10% on the month before · 72 mentions on Trending in 14 days
Models split into many expert blocks with only a few active per token, so they grow in size without growing in cost. Named by Shazeer et al., Google Brain: Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer (Jan 23, 2017). Also called MoE, Sparse models, Sparse mixture of experts.
Index 0 to 100, monthly
Signals
Hacker News stories, last 12 months110 stories, +244% on the 12 months before
arXiv papers, last 12 months1,998 papers, +84% on the 12 months before
Wikipedia views, October 20268,611, +4% on a year beforeProduct releases naming it17 since Feb 19, 2026GitHub repositories2,012| Signal | Reads as | |
|---|---|---|
| Hacker News stories, last 12 months110 stories, +244% on the 12 months before | 110 stories, +244% on the 12 months before | Adoption |
| arXiv papers, last 12 months1,998 papers, +84% on the 12 months before | 1,998 papers, +84% on the 12 months before | Adoption |
| Wikipedia views, October 20268,611, +4% on a year before | 8,611, +4% on a year before | Adoption |
| Product releases naming it17 since Feb 19, 2026 | 17 since Feb 19, 2026 | Adoption |
| GitHub repositories2,012 | 2,012 | Adoption |
Companies
| Name | Mentions |
|---|---|
| 6 | |
| 3 | |
| 3 | |
| 3 | |
| 1 | |
| 1 |
Releases
17 releases since Feb 19, 2026| Name | |
|---|---|
| Oct 29, 202611 days ago | |
| Oct 19, 20263 weeks ago | |
| Oct 15, 20263 weeks ago | |
| Oct 14, 20263 weeks ago | |
| Sep 10, 20262 months ago | |
| Sep 9, 20262 months ago | |
| Aug 29, 20262 months ago | |
| Aug 28, 20262 months ago |
Models
DeepSeek V3DeepSeekDeepSeek V3.1DeepSeekQwen3 235b A22B Diff No AlibabaAlibabaQwen3-235B-A22B-Thinking (Jul 2025)AlibabaLlama4 ScoutMetagpt-oss-120bOpenAI
| Name |
|---|
| DeepSeek V3DeepSeek |
| DeepSeek V3.1DeepSeek |
| Qwen3 235b A22B Diff No AlibabaAlibaba |
| Qwen3-235B-A22B-Thinking (Jul 2025)Alibaba |
| Llama4 ScoutMeta |
| gpt-oss-120bOpenAI |
Papers
GitHub projects
2,012 repositoriesdeepseek-ai/DeepSeek-VL2DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understandingdeepseek-ai/DeepSeek-V2DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language ModelPKU-YuanGroup/MoE-LLaVA【TMM 2025🔥】 Mixture-of-Experts for Large Vision-Language Modelsdeepseek-ai/DeepSeek-MoEDeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language ModelsXueFuzhao/OpenMoEA family of open-sourced Mixture-of-Experts (MoE) Large Language ModelsXueFuzhao/awesome-mixture-of-expertsA collection of AWESOME things about mixture-of-experts
| Name | Stars | |
|---|---|---|
| deepseek-ai/DeepSeek-VL2DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding · Dec 13, 2024 1 year ago | Dec 13, 20241 year ago | 5,375 |
| deepseek-ai/DeepSeek-V2DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model · Apr 22, 2024 2 years ago | Apr 22, 20242 years ago | 5,041 |
| PKU-YuanGroup/MoE-LLaVA【TMM 2025🔥】 Mixture-of-Experts for Large Vision-Language Models · Dec 14, 2023 2 years ago | Dec 14, 20232 years ago | 2,324 |
| deepseek-ai/DeepSeek-MoEDeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models · Jan 2, 2024 2 years ago | Jan 2, 20242 years ago | 1,976 |
| XueFuzhao/OpenMoEA family of open-sourced Mixture-of-Experts (MoE) Large Language Models · Aug 8, 2023 3 years ago | Aug 8, 20233 years ago | 1,699 |
| XueFuzhao/awesome-mixture-of-expertsA collection of AWESOME things about mixture-of-experts · Mar 30, 2022 4 years ago | Mar 30, 20224 years ago | 1,287 |
Related movements
Open-weight modelsModels whose trained weights anyone can download, run and fine-tune: Llama, Qwen, DeepSeek, Mistral and more.Inference chipsAccelerators built to serve models cheaply and fast rather than to train them: TPUs, LPUs, wafer-scale and custom silicon.Reasoning modelsModels trained with reinforcement learning to think step by step, spending more compute at answer time.
| Movement | Stage |
|---|---|
| Open-weight modelsModels whose trained weights anyone can download, run and fine-tune: Llama, Qwen, DeepSeek, Mistral and more. | Rising |
| Inference chipsAccelerators built to serve models cheaply and fast rather than to train them: TPUs, LPUs, wafer-scale and custom silicon. | Peak |
| Reasoning modelsModels trained with reinforcement learning to think step by step, spending more compute at answer time. | Peak |
History
Tracking started: emerged January 2017, peak, 3 other names recorded
| Date | What changed |
|---|---|
| Sep 24, 2026 | Tracking started: emerged January 2017, peak, 3 other names recorded |
Sources: Curve: Hacker News story titles, Wikipedia pageviews and arXiv papers by month, refreshed Nov 1, 2026. Related entities from the other Glyph Ibis sources' public APIs, matched by name. How stages work.