Trend tracker
Open-Source MoE Competition
Open MoE models like the DeepSeek V4 series, Qwen and Kimi K3 launch in rapid succession, matching closed flagships — while Meta retreats from open source: its Llama API public preview shut down in July 2026 and the never-released 'Llama 5' gave way to the closed Muse line.
Editorial heat index
179
editorial cumulative index
Tracking period
2026 YTD
Coverage in this tracker
12 tracked stories
The heat index is a cumulative editorial index AI Tools Daily maintains, scored on a 20–400 scale. It is a running score — not a percentage, a growth rate or a market share, and not third-party statistics.
How the heat index moves Each day the refresh script reads the site's nine English news feeds (OpenAI, Google, Google DeepMind, TechCrunch AI, The Verge AI, Ars Technica AI, MIT Technology Review, Hugging Face, NVIDIA) and samples titles published in the last 7 days, up to 20 per feed. A title counts as one mention when its text contains any of the tracker's ASCII tags — tags matching ^[a-z0-9 .-]{3,}$; Chinese-language tags are filtered out and never match. A tracker with mentions gains min(mentions, 5) points, one with none loses 1 point, and the score always stays within 20–400. If every feed fails or no title is sampled, the score is left unchanged that day.
Method note updated 2026-10-14
What this tracker follows
The coverage list below matches these keywords against story tags, headlines and standfirsts in the archive.
Coverage in this tracker
- · 2 min readQwen 3 Released: MoE Architecture Sets a New Bar for Open Source
- · 2 min readKimi K2.5 Released: Trillion-Parameter MoE Trained on 15T Mixed Visual-Text Tokens
- · 2 min readMoonshot AI Releases Kimi K2: A Trillion-Parameter MoE Model, Open-Sourced
- · 1 min readMeta Releases Llama 4: Scout and Maverick Debut as Natively Multimodal Open Models
- · 1 min readOpen-Source LLM Competition Heats Up: Llama 4, Mistral Large 2 and More
- · 2 min readDeepSeek V3 Released: A 671B MoE Model That Shook the Industry on Cost
- · 2 min readQwen3-Max Released: Alibaba's First Trillion-Parameter Model, Pretrained on 36T Tokens
- · 2 min readOpenAI Returns to Open Source After Six Years: gpt-oss Ships Under Apache 2.0
- · 2 min readDeepSeek V4.1 Flash Launches—Reconstructed From Third-Party Trackers
- · 1 min readOpen vs Closed LLMs: Four Dimensions for Enterprise Decision-Making
- · 2 min readOpen Weights and Chinese Models Narrowed the Frontier Gap to About 4.4 Months
- · 3 min readAnthropic's Dario Amodei Responds: Never Advocated Banning Open-Weight Models — the Real Fear Is Chinese AI