By olud.ai editorial · built from our own tracking data, published daily
DeepSeek V4 Pro 0423 input price fell 22% to $0.42 per 1M tokens, leading a wave of price reductions across four major models. GLM 5.1 from Z.AI dropped 23% to $0.97, Qwen3 Next 80B A3B Instruct from Alibaba fell 10% to $0.09, and Devstral 2 2512 from Mistral AI declined 9% to $0.4 per 1M tokens.
The ecosystem gained 57,684 GitHub stars across 10,000+ tracked open-source AI projects today, with 66 new projects entering the index. Weekly totals show 1,872 new projects and 100 new Hugging Face spaces added. The top gainer was archify with 3,433 stars.
Two models saw price increases: Inkling from Thinkingmachines rose 5% to $1 per 1M tokens, and Trinity Large Thinking from Arcee ai increased 14% to $0.25 per 1M tokens.
Releases of the day
Four minor releases shipped today. n8n 2.36.9 fixes a core bug applying proxy environment variables consistently across packages and processes. Open WebUI v0.11.2 adds richer previews for terminal files, showing Word documents and slide decks as finished documents. Sentence Transformers v6.0.1 patches a multi-vector loading bug where PyLate checkpoints with [Q]/[D] prefixes and text prompts lost the prefix during encoding. LanceDB v0.38.0 introduces breaking changes, making table existence manifest-authoritative and requiring pydantic v2 in Python.
▼ 10%Qwen3 Next 80B A3B InstructAlibaba · input $0.1 → $0.09 per 1M tokens
▼ 9%Devstral 2 2512Mistral AI · input $0.44 → $0.4 per 1M tokens
▲ 5%InklingThinkingmachines · input $0.95 → $1 per 1M tokens
▲ 14%Trinity Large ThinkingArcee ai · input $0.22 → $0.25 per 1M tokens
Emerging projects
Three new projects entered the index today. Atlas is a Rust project with 1,416 stars providing source control for agents, allowing users to manage multiple coding agents and track their changes. Locally-uncensored is a TypeScript project with 1,301 stars offering a plug-and-play local AI studio with uncensored chat, image and video generation, and a coding agent running abliterated LLMs plus ComfyUI fully offline. Agent-qa is a TypeScript project with 896 stars providing an open-source self-improving QA agent with memory for software teams, supporting natural language tests for web and mobile.
A Hugging Face community paper with 35 upvotes titled "Agentic Artifact Creation: Systems, Evaluation, Principles, and Opportunities" examines how generative models turn natural-language prompts into images, text, and code. The authors argue that practical impact depends on whether these pieces can become complete artifacts.
Today's widespread price cuts, especially the 22% drop from DeepSeek and 23% from GLM 5.1, signal continued downward pressure on inference costs that benefits developers and small teams. The emergence of agent-focused tools like atlas and agent-qa, combined with the research focus on artifact completion, suggests the ecosystem is maturing from raw generation toward integrated agent workflows.
Source: olud.ai tracking of 10,000+ open-source AI projects, 300+ models and live provider pricing. Figures are measured, not estimated. All releases · Live pricing · Latest in AI
