The artificial intelligence market has entered a phase of intense competition for speed. On August 13, both tech giants simultaneously unveiled solutions aimed at radically accelerating work with code and autonomous AI agents. Google announced the Gemini 3.7 Flash model, while OpenAI introduced an experimental Ultrafast mode for GPT-5.6 Sol. These are not just updates, but a marker of shifting priorities: the battle is now not only over reasoning quality, but also over milliseconds of response time.
What's new in Gemini 3.7 Flash
Google positions Gemini 3.7 Flash as a specialized tool for programming and business process automation. Key improvements concern multi-stage planning, accuracy in following instructions, and code generation. However, the most important signal for the market was the twofold reduction in usage cost compared to its predecessor. This is a direct blow to competitors' positions in the price segment.

The Flash line has historically been created for scenarios where low latency is critical: coding, AI agent work, and tasks with many sequential calls to the model. Recall that in July, Gemini 3.6 Flash was released with a price of $1.5 per 1 million input tokens and $7.5 per 1 million output tokens, which already demonstrated 17% greater efficiency in using output tokens than version 3.5 Flash. The new model continues this trend.
Notably, the flagship Gemini 3.5 Pro has still not been released — developers are testing it with partners, and the release timeline remains unclear. This creates intrigue: perhaps Google has decided to bet on speed and cost-effectiveness rather than maximum power.
How OpenAI accelerated GPT-5.6 Sol
OpenAI responded symmetrically, but with a different emphasis. The Ultrafast mode for GPT-5.6 Sol promises up to 14x acceleration compared to the standard mode, delivering up to 750 output tokens per second. This is achieved through the use of Cerebras infrastructure — specialized chips optimized for high throughput.
At launch, Ultrafast is available via API only to a limited circle of clients. This is a classic phased rollout strategy: first, a stress test on early users, then scaling as computing capacity grows. The figure of 750 tokens per second was mentioned back in the June announcement of GPT-5.6, but at that time developers honestly warned that the accelerated version would not be available to everyone at once.
In this context, it is worth recalling the August release of Grok 4.6 from SpaceXAI, focused on long-term agentic tasks. The market is clearly moving toward specialization: speed is becoming as important a parameter as the quality of logic.
My analysis: The dual announcement on the same day is not a coincidence, but a deliberate escalation. We are witnessing the formation of a new standard where response latency becomes a first-order competitive advantage. For Web3 and DeFi protocol developers, this means the ability to create more responsive agents, but also growing demands on infrastructure. Keep an eye on prices: Google's dumping could provoke a reciprocal tariff reduction from OpenAI, which ultimately benefits all ecosystem participants.