Alibaba’s Qwen team released open-weight models under the Apache 2.0 license with Qwen 3.8, a dense 27-billion-parameter model designed to outperform the larger Qwen 3.7 Plus in coding and office tasks while natively processing up to 262,000 tokens of context (The Decoder). Separately, Zhipu AI released GLM-5.3, which according to its own benchmarks achieves a 50 percent improvement over its predecessor through post-training alone, reusing the 743B GLM-5.2 base model unchanged. Terminal-Bench 3.0 scores moved from 4.6 to 28.3, and DeepSWE v1.1 from 46.2 to 66.9, with the model trained for cybersecurity and claimed to have helped security teams identify 2,436 vulnerabilities (The Decoder, MarkTechPost).
OpenAI launched Ultrafast, a new inference mode delivering GPT-5.6 Sol at up to 750 output tokens per second, powered by Cerebras hardware from their $10 billion partnership, creating a three-tier pricing structure that turns inference speed into a cost lever (The Decoder). Cactus Compute released Needle 2, an open 45-million-parameter tool-calling model shipped as a single 14MB binary that runs a full session in approximately 28MB of RAM, targeting hardware with no GPU or NPU (MarkTechPost).