A developer ran a 90-million-parameter conversational language model entirely on a Sony PSP, the gaming handheld released in 2004. The LLMPSP project on GitHub shows that 90M parameters is about the practical ceiling for the device before response quality degrades significantly. It serves as a fun benchmark for just how constrained early 2000s mobile hardware really is.
/5 Sep 2026/Rreddit

indextkn/skills is a set of agent skills that provide live pricing data for over 900 AI models across 15 API providers. It runs as an agent tool, letting developers query current inference costs without leaving their workflow. The data covers all major commercial and open-weight model providers and is updated in real time.
/30 Aug 2026/GGitHub

energygraph v1.3 adds discrete GPU power monitoring for NVIDIA, Intel, and AMD cards, extending the tool beyond its previous CPU-only scope. The tool is open-source with zero build dependencies and displays live power consumption data directly in the terminal.
/23 Aug 2026/Rreddit

nRouter SDK puts a single API key and one interface in front of six cloud model providers: Alibaba US, OpenAI, AWS Bedrock, Azure Foundry, Google Vertex AI, and Anthropic. It speaks both the OpenAI and Anthropic wire formats natively and covers chat completions, embeddings, images, text to speech, speech to text, and video. Branded packages ship for Python, JavaScript, Java, Rust, and R, with OpenAI-compatible paths for Go, Ruby, PHP, and C#. Framework integrations include LangChain, LlamaIndex, Vercel AI SDK, CrewAI, and AutoGen.
/22 Aug 2026/GGitHub