News Hub

Baseten released Foundations, a platform for inference, training, and execution.

Baseten Research developed legal agents through post-training methods to handle frontier legal tasks.

Baseten describes infrastructure as the platform enabling inference, training, and execution of AI models.

Baseten and Postman collaborate on building reliable, cost-efficient agents for enterprise use.

Baseten releases joey-1, a fast decision model optimized for low-cost inference.

Qwen3-ASR provides fast, low-cost batch audio transcription.

Baseten announced new engineering leaders.

Baseten introduced the Environment Network.

Beam is an efficient reasoning system for coding and agentic workloads.

Article discusses characteristics that define high-quality NVFP4 quantization.

Optimization techniques achieve 50-90% faster inference speeds for agentic AI systems.

Baseten announced a partnership with OpenAI.

Baseten introduced hosted tools for deploying and managing AI models.

DeepSeek-V4.1-Flash improves prefill efficiency for coding agents.

Blaxel is joining Baseten to build agentic infrastructure.

Baseten leads a benchmark for voice AI systems.

Baseten released a new MCP and skill enabling coding agents to use Baseten's platform.

Article discusses optimization techniques for reducing computational costs during language model inference.

Agentic kernels are now deployed in production systems.

GLM 5.3 scales model performance through post-training techniques.