Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A developer deployed a Llama 3 inference endpoint on Oracle Cloud Infrastructure's OKE cluster using vLLM and NVIDIA A10 GPUs in about 20 minutes. The setup cost $1.52/hr on-demand or $0.46/hr preemptible, significantly …
A developer introduced Graphify Companion, an open-source tool that builds a local 3D codebase knowledge graph and syncs LLM context offline. The tool combines Node.js for MCP Server Transport and Python for AST/graph pa…
The U.S. government ordered Anthropic to restrict access to its Fable 5 AI model over national security concerns, citing partial jailbreaking that could enable dangerous uses like designing pathogens. Anthropic disputes …
Flama 2.0 introduces first-class support for generative AI, enabling users to download, package, and serve large language models (LLMs) via a single command line. The framework allows fetching models from HuggingFace, in…
Oxlo.ai's request-based pricing enables developers to build a conversational language tutor that corrects mistakes in real time and adapts to learner proficiency. Using the llama-3.3-70b model via the OpenAI SDK, the sys…
A freelance developer rebuilt their AI stack around Airtable AI, reducing monthly API costs from $89 to $14—an 84% drop—by switching from GPT-4o to cheaper models like DeepSeek V4 Flash and Qwen3-32B via a global API gat…
Oxlo.ai offers request-based pricing for LLM-driven named entity recognition, making it economically viable for long documents. The platform supports structured output via JSON mode and function calling, enabling flexibl…
A developer shipped a small research agent that addresses three common LLM limitations: stale training data, hallucinated facts, and arithmetic errors. The agent uses tool calling to look up facts and verify math before …
A developer built a Python CLI tool that uses Oxlo.ai's LLM to evaluate whether a business task is suitable for automation with a large language model. The tool sends a task description to the llama-3.3-70b model and ret…
Microsoft is considering a fine-tuned version of DeepSeek V4 as a cheaper model for Copilot Cowork and switching to usage-based billing, as flat-rate pricing is unsustainable, according to Copilot head Charles Lamanna.
Oxlo.ai offers flat per-request pricing for LLM APIs, decoupling cost from context size and enabling long-context applications without token-based billing. The company provides techniques such as prompt compression, mode…
Oxlo.ai is building an autonomous research agent that converts vague questions into structured plans, gathers evidence across multiple LLM calls, and synthesizes markdown reports. The agent uses small, orchestrated reaso…
Oxlo.ai has introduced a request-based pricing model and an OpenAI-compatible API supporting over 45 models across seven categories, addressing the cost unpredictability of token-based billing for long-context and agenti…
Oxlo.ai offers request-based pricing for AI inference, charging a flat fee per API call regardless of prompt length, contrasting with token-based models used by providers like Together AI and Fireworks AI. This approach …
A developer compared LLM inference APIs on cost, performance, and integration, noting that most providers use token-based pricing which can cause unpredictable costs for long-context or agentic workloads. Oxlo.ai offers …
Meta launched AI Mode in Facebook Search, using Meta AI to generate answers from public content in Groups, Reels, and other apps. The feature aims to surface real experiences and recommendations, but Meta did not explain…
Chief Judge Martin Reidlinger of the U.S. District Court for the Western District of North Carolina declined to sanction plaintiff's counsel after they admitted to using artificial intelligence tools that generated hallu…
Microsoft is moving its Copilot Cowork agent to usage-based pricing and exploring a fine-tuned version of DeepSeek V4 or another open-source model as a lower-cost alternative, Axios reports. The optional model will be ho…
A developer built a custom AI agent evaluation harness in a single Bun file using the Claude CLI, avoiding external frameworks or SaaS platforms. The harness runs the agent in a sandbox, grades outputs with string checks…
Anthropic enabled subagents to launch their own subagents, leading to reports that ed3d-plugins are costing users hundreds of dollars in extra API tokens due to uncontrolled subagent creation.