NEWS / 01
Follow recent records from AI models, tools, and companies by publication time.
Latest release
Reliance Jio collaborates with Google to offer eligible users 18 months of free access to the Google AI Pro subscription, which includes Gemini Advanced and 2TB cloud storage.
Timeline
Alibaba's Qwen team launched Qwen3.8-Max-Preview on July 19 through Token Plan, Qoder, and QoderWork. The company says the Qwen3.8 family reaches 2.4 trillion parameters and will soon release open weights, while full benchmarks, a model card, licensing terms, and a weight-release date remain undisclosed.
The 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance opened in Shanghai on July 17. Representatives of 29 countries signed the agreement establishing the World Artificial Intelligence Cooperation Organization, which will be headquartered in Shanghai; organizers also announced plans for more than 140 forums, over 1,100 exhibitors, and more than 300 global product debuts. The event was still underway as of July 18, so final launch and partnership figures remain subject to the post-conference report.
Kimi K3 leads seven of Moonshot AI's 35 published comparisons and jointly leads one more, while Arena ranks it first for frontend development and ninth in text overall. For a workload using 1M input and 200K output tokens with no cache hits, K3 costs about $6—70% less than Claude Fable 5 and roughly 45% less than GPT-5.6 Sol, but about 2.63 times the listed cost of GLM-5.2.
Moonshot AI launched its flagship Kimi K3 model on July 16, 2026. The 2.8-trillion-parameter MoE model natively supports vision and a 1-million-token context window. It is live across Kimi, Kimi Work, Kimi Code, and the API, with full model weights scheduled for release by July 27.
As the marginal returns of model raw performance decline, the bottleneck of enterprise AI shifts to deployment. In May 2026, OpenAI launched a $4 billion deployment company, Anthropic embedded FDEs within financial giant FIS, and Google Cloud initiated a massive hiring drive for FDEs. The Forward Deployed Engineer (FDE) model, pioneered by Palantir, is becoming the key strategy for tech giants to secure enterprise workflow entries.
Alibaba Cloud officially launched Qwen-Audio-3.0-Realtime, a speech interaction model designed for low-latency, real-time voice applications. It features millisecond-level responsiveness, full-duplex interruption, and proactive Agent tool execution, offering Plus and Flash variants.
The highly anticipated Google Gemini 3.5 Pro has faced multiple delays, with the latest timeline pointing between July 17 and early August. Early tests on LMArena revealed token exhaustion issues, pushing Google to retrain the model.
In July 2026, developer @Lonely__MH warned that using the Superpowers skill with the GPT-5.6 Sol model causes severe conflicts. The legacy framework clashes with Sol's native reasoning, leading to context pollution, token explosions, and dead loops.
In-depth analysis of the latest DeepSew benchmark data: Under the premise of meeting the performance baseline, using absolute low price as the highest standard, revealing gpt-5.6-luna as the true king of cost-effectiveness today.