AI's model race stayed loud over the past day: Google shipped a cheaper, faster Gemini, DeepSeek and Z.ai pushed out new models with fresh limits attached, and Anthropic and Apple made moves off the model page—one a possible $6 billion deal, the other a homegrown model for China. Here's the rundown →

TOGETHER WITH OUR SPONSOR

Creators lock in their holiday content calendars 90 days out, before most ecommerce brands finalize their commission strategy and way before Black Friday and October deal events.

Get ahead of the seasonal rush with The 90-Day Holiday Sprint, a practical guide for brands that want creators driving holiday demand while competitors are still recruiting:

Your 90-day countdown starts now.

Get the Free Guide

Google releases Gemini 3.7 Flash for coding and agent work. Google introduced Gemini 3.7 Flash, positioning it as its most capable “workhorse” model for software engineering and agent workflows, roughly three weeks after Gemini 3.6 Flash. The company says the model improves on debugging, issue resolution, and first-pass code accuracy, and reports that it outperformed comparable models from Anthropic and OpenAI on nine benchmarks. Introductory API pricing is $0.75 per million input tokens and $3.75 per million output tokens—about half the prior Flash rate—through December 31, 2026, after which it doubles. The release arrives while Google's higher-end Gemini 3.5 Pro remains delayed. Source

DeepSeek moves V4-Pro out of preview and flags a price increase. DeepSeek made its flagship V4-Pro model generally available under the build name V4-Pro-0813, with a one-million-token context window, up to 384,000 output tokens, and MIT-licensed weights for self-hosting. The company said its current rates—about $0.44 per million input tokens on a cache miss and $0.87 per million output—hold only until 16:00 UTC on August 16, when both peak and off-peak tiers rise. Even after the increase, DeepSeek's published pricing remains far below that of leading US frontier models. Source

Z.ai launches GLM-5.3 but holds back the weights. Chinese lab Z.ai (formerly Zhipu) released GLM-5.3, which it says keeps the same 744-billion-parameter base as GLM-5.2 and derives its gains entirely from expanded post-training. The company reports it as its strongest open-weights system for coding and says its cybersecurity ability grew faster than expected during training. GLM-5.3 is available now through Z.ai's API and coding plan, but the weights will be published in about two weeks pending safety hardening, and more sensitive capabilities are gated behind a “Cybersecurity Trusted Access” verification program. Source

OpenAI previews an “Ultrafast” mode running GPT-5.6 Sol up to 14x faster. OpenAI began previewing Ultrafast mode, which it says runs its GPT-5.6 Sol model at up to 14 times the speed of standard processing and generates up to roughly 750 output tokens per second. The mode is powered by Cerebras hardware through the two companies' low-latency inference partnership and is launching first in the OpenAI API to a limited group of customers testing coding, commerce, financial research, and support workloads. OpenAI has not published pricing or a general-availability date. Source

Anthropic is reported to be in talks to buy Decart for about $6 billion. Anthropic is in discussions to acquire Israeli startup Decart in a deal valued near $6 billion, which would be its largest known acquisition, according to reports; the talks are ongoing and could still fall through. Decart builds models beyond text, including Lucy, which edits live video in real time, and Oasis, which generates simulated environments for training robotics and autonomous-driving systems. The startup raised $300 million in May at a valuation near $4 billion in a round led by Radical Ventures, with Nvidia participating. Source

Apple builds a China-specific AI model with Alibaba. Apple has trained its own large language model for the Chinese market with help from Alibaba, a shift from its earlier reliance on outside partners' models there, according to Reuters. Reports describe Apple as the first foreign company cleared by Chinese regulators to offer a proprietary generative-AI model in the country, with Alibaba's Qwen integrated into Apple Intelligence for Chinese users across its operating systems. Apple Intelligence is expected to reach Chinese iPhones in the coming months following an iOS update. Source

TOGETHER WITH OUR SPONSOR

Tax season doesn't have to mean wondering if you have the right forms, second-guessing your deductions, or scrambling to pull everything together before the deadline.

With BELAY’s Tax Prep Checklist, you can start preparing for tax season with confidence.

Download the Free Checklist

That's today's AI Current. Thanks for reading—reply and tell us what you'd like us to cover next. See you tomorrow.

TOGETHER WITH OUR SPONSOR

Opening entities in every country can be slow, expensive, and hard to scale.

That's why more companies are using EOR to hire globally faster.

See how Oyster helps teams hire, pay, and support talent in 180+ countries while staying compliant along the way.

Get all the answers