Google released Gemini 3.7 Flash, arriving just three weeks after Gemini 3.6 Flash with claimed improvements in coding and agentic workloads. According to Google’s benchmarks, the model outperforms Claude Sonnet 5 and GPT-5.6 Terra while pricing at 50 percent below its predecessor, positioning it as a cost-efficient option for agent deployment and development tasks (The Decoder). Separately, DeepSeek moved its V4-Pro model out of testing and released Harness v0.1, its agent orchestration software, under the MIT open-source license (The Decoder). DeepSeek also raised API pricing, with cache hits increasing to six times their current cost, which affects cost economics for agents that repeatedly access the same data.
Writer introduced a new model built as a post-training variant on Z.ai’s open-source GLM-5.2, emphasizing lower token costs and deployment-ready capabilities as a cost-containment strategy (TechCrunch). OpenAI launched Ultrafast, a preview mode for GPT-5.6 Sol that achieves 14x speedup, targeting enterprise users concerned with inference latency (TechCrunch).