Input prices jump up to 275% as agent tooling ships updates
86,035 stars today265 new projects+51,584 rea
By olud.ai editorial · built from our own tracking data, published daily
Input prices moved sharply across the tracked model catalog on October 14, 2026. OpenAI's gpt-oss-120b rose 275% to $0.15 per 1M tokens. DeepSeek V4 Flash 0423 rose 200% to $0.09 per 1M tokens. Meta's Llama 3.3 70B Instruct rose 168% to $0.59 per 1M tokens.
The broader ecosystem gained 86,035 GitHub stars across the 10,000+ tracked open-source AI projects so far today. A total of 265 new projects entered the index. The top gainer, "rea", added +51,584 stars today.
The weekly context shows 2,010 new projects entering the index this week, plus 100 new Hugging Face spaces. Several minor releases also landed across agent frameworks, workflow tools and model gateways.
Releases of the day
Hermes Agent v0.22.0 shipped with roughly 24,200 commits, 6,080 merged PRs and about 2,045,000 insertions since v0.21.0. n8n 2.43.5 fixed validation of the chunk() expression extension and added Gemini models by name. LiteLLM v1.105.1 moved to cosign-signed Docker images for every release. CrewAI 1.15.28 added a crewai eval command and switched to the LiteLLM catalog for context windows. Skyvern v1.0.56 now sends the current screenshot on every Copilot call. SurfSense v2.2.1 added plugin support for agentic chats, covering Google Search, Web Crawl, YouTube, TikTok and Google Maps. Milvus v3.0.3 and ArduPilot Plane-4.7.2 also shipped updates.
Six models saw input price increases and two saw decreases. OpenAI's gpt-oss-120b rose 275% to $0.15 per 1M tokens, the largest move. DeepSeek V4 Flash 0423 rose 200% to $0.09, Llama 3.3 70B Instruct rose 168% to $0.59, Qwen3.5-35B-A3B rose 87% to $0.15, and MiniMax M2.7 rose 43% to $0.3 per 1M tokens. GLM 5.3 Flash fell 80% to $0.03 per 1M tokens. DeepSeek V4.1 Flash fell 73% to $0.04, and DeepSeek V4 Pro 0423 fell 2% to $0.4 per 1M tokens. Developers using these models via the live pricing index will see the changes immediately.
The index added 265 new projects today, following 2,010 new entries earlier this week. One hundred new Hugging Face spaces also appeared this week. No new project names were disclosed beyond the day's top gainer, "rea".
Research pick
The Hugging Face community pick is "Vela: User Behavior as Supervision Sparks LLM Intelligence" by Yang Xiao, Yusong Sun, Haoyi Wu and colleagues. It argues that many LLM applications, such as personal assistants and recommendation systems, depend on long-term user behavior histories. The paper proposes using that user dimension as a supervision signal to improve LLM intelligence.
The day's facts point to two parallel trends: model input prices are volatile, with sharp increases on several popular models, while agent and workflow tooling continues to ship steadily at a minor-release pace. The large star gain for a single project also signals concentrated community attention in the open-source index.
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
