News Hub

CWE-bench v2 provides expanded coverage, improved quality, and greater transparency for security benchmarking.

Qwen4 models and release date announced.

Title alone insufficient for concrete summary.

FlexAI provides five checks for evaluating inference providers.

FlexAI identifies five checks for selecting open-source LLMs with function calling capability.

Article discusses when to switch between open source and closed LLMs.

EasyR1 RL fine-tuning compares GRPO versus DAPO methods.

Article about methods for benchmarking open-source language models.

A 32 billion parameter legal language model achieves frontier performance at four times lower cost.

Article about selecting appropriate GPUs for language model inference workloads.

FlexAI released FlexBench, an open-source MLPerf LLM inference benchmark.