By olud.ai editorial · built from our own tracking data, published daily
DeepSeek cut input pricing for V4 Flash 0423 by 67 percent to $0.01 per 1M tokens on October 9, while raising V4 Pro 0423 input price by 38 percent to $0.29. The moves make the Flash model markedly cheaper for high-volume input workloads.
The olud.ai index gained 66,572 GitHub stars across the 10,000-plus tracked open-source AI projects so far today, and 61 new projects entered the index. This week, 1,872 new projects entered, plus 100 new Hugging Face spaces.
The release track was light. n8n shipped 2.42.6 and Phoenix shipped 20.20.0, both as minor updates focused on workflow control and evaluation tooling.
Releases of the day
n8n 2.42.6 blocks creation and updates of workflows that contain deprecated nodes, and the Google Gemini node now shows Nano Banana 2.1 and 2 Lite options. Phoenix 20.20.0 adds a prompt hill-climb task for a text-to-SQL dataset and allowlists annotation, evaluator, prompt, and split in its mcp/sql integration. These changes tighten guardrails in workflow automation and expand evaluation options for agent and SQL pipelines. Users of both tools get incremental control and visibility rather than a major feature shift.
DeepSeek V4 Flash 0423 input price dropped 67 percent to $0.01 per 1M tokens, while DeepSeek V4 Pro 0423 rose 38 percent to $0.29. GLM 5.2 fell 59 percent to $0.07, Kimi K2.6 fell 7 percent to $0.43, Qwen3 Next 80B A3B Instruct fell 10 percent to $0.09, and Llama 3.3 70B Instruct fell 55 percent to $0.1. On the upside, Mistral Nemo rose 50 percent to $0.03 and Nemotron 3 Nano 30B A3B rose 20 percent to $0.06.
The research pick is "Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments?" by Yibo Li, Jinhang Qiu, Zhi Zheng et al., shared on Hugging Face with 91 upvotes. It focuses on evaluating whether LLM agents can learn from experience and adapt to unfamiliar, dynamic environments, an area existing benchmarks do not fully cover.
Today's price changes split by model tier, with cheaper flash and smaller models alongside more expensive flagship and niche options. The ecosystem keeps adding projects steadily, and the new tools cluster around agent oversight, coding workflows, and execution safety.
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
