AI Intel

Ollama's v0.40.0-rc0 release now defaults to running supported model architectures on MLX for Apple Silicon devices, enhancing performance.

Applying a logit penalty to hesitation words like "wait," "maybe," and "perhaps" significantly enhances the accuracy of Qwen models, as shown in test…

A new offering named "GPT-6 Sol & Luna" is advertised on Product Hunt as providing frontier AI intelligence at a 50% reduced price.

Harmony introduces AI agents designed to automate the resolution of IT and HR tickets directly within popular collaboration platforms like Slack and…

Humalike x GTA RP showcases AI NPCs capable of independent conversation, memory retention, and autonomous actions, enhancing role-playing experiences.

DeepSeek-V4-Pro API is offering a 75% discount until May 2026, with integration updates enabling Claude Code to access its 1M context window.

Agent Arena's analysis of top models reveals Kimi K3 (Max) delivers strong performance at a median cost of $0.62 per task, significantly cheaper than…

Agent Arena has introduced a Pareto frontier analysis, comparing model performance on real-world agentic tasks against their median cost, featuring C…

vLLM v0.30.0 is released, featuring 762 commits and new model support including DeepSeek-V4.1-Flash with MXFP8 KV storage, particularly optimized for…

Hugging Face announced SmolDataEnvs, an open-source collection of 5,000 verifiable RL environment tasks tailored for training and evaluating small mo…