AI ecosystem sees 172k stars, DeepSeek V4 Pro price jumps 95%
172,501 stars today74 new projects+51,855 rea
By olud.ai editorial · built from our own tracking data, published daily
The open-source AI ecosystem gained 172,501 GitHub stars today across 10,000+ tracked projects, with 74 new projects entering the index. The top gainer "rea" added 51,855 stars alone.
This week 1,787 new projects entered the index, alongside 100 new Hugging Face spaces. The pace of new entries remains high, reflecting continued experimentation across agent systems, video generation, and inference engines.
Three releases shipped: Langfuse v4.58.0 adds Turnstile CAPTCHA to password-reset emails; Weaviate v1.40.1 fixes a 500 error on object ID checks; Phoenix arize-phoenix-sqlean v0.2.0 bundles SQLite 3.54.0 and supports Python 3.15. Meanwhile, several model prices moved sharply, with DeepSeek V4 Pro leading increases.
Releases of the day
Langfuse v4.58.0 introduces Turnstile verification on password-reset emails, a security-focused feature. Weaviate v1.40.1 contains no breaking changes or new features, only a fix that returns a 500 status instead of an "already exists" message when checking an object ID fails. Phoenix arize-phoenix-sqlean v0.2.0 upgrades its bundled SQLite to 3.54.0 and adds Python 3.15 support, keeping the observability library compatible with the latest runtime.
DeepSeek V4 Pro 0423 input price rose 95% to $0.41 per 1M tokens, the largest increase today. Llama 4 Scout input price jumped 80% to $0.18 per 1M tokens. Gemma 4 26B A4B and GLM 5.3 each rose 14% to $0.08 per 1M tokens. On the downside, DeepSeek V4.1 Flash input price dropped 50% to $0.15 per 1M tokens, and Kimi K3 fell 9% to $0.73 per 1M tokens. The moves affect users of these models on API platforms that track live pricing.
▲ 95%DeepSeek V4 Pro 0423DeepSeek · input $0.21 → $0.41 per 1M tokens
Seven new projects were detected today. "dsh-our-free-model" (8,836 stars) is a JavaScript plugin that provides free, unlimited access to models including DeepSeek V4.1 Flash and Kimi K3 without login or API key. "reelmimic" (2,209 stars) generates a new video in the style of a reference video, using Claude Code or Codex as an AI crew. "lemo-opuscar" (1,896 stars) is a Claude Code skill for short films with 43 film styles, each with a style prompt and a demo film made entirely in code. "answer-me-with-html" (3,106 stars) is an agent skill that answers hard questions with a single readable HTML page. "hello-agent-system" (146 stars) offers 17 bilingual lessons on building enterprise AI agent systems from scratch. "monid" (3,673 stars) positions itself as "OpenRouter for agent tools." "Misaka-Agent" (196 stars) is a multi-agent research system for humanities and social sciences. "TensorFold" (1,255 stars, written in Zig) is an LLM inference engine supporting Metal, CUDA, and Vulkan.
The Hugging Face community pick "GroundingPI" (91 upvotes) proposes a grounding foundation model for physical intelligence. It uses visual primitives to specify which object is the target and where it is, even in clutter and for tiny objects, and is designed to be fast enough for closed-loop control. The work addresses a key limitation in vision-language-action models and world-action models.
Today's facts show a market splitting: model prices diverge sharply while the open-source ecosystem continues to absorb new projects at a high weekly rate. The emergence of free-access plugins and agent skills suggests users are seeking ways to bypass rising API costs, and the research focus on precise grounding points toward the next frontier in embodied AI.
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
