Google and OpenAI have launched new AI models focused on high-speed execution for coding and automation tasks.
On August 13, Google announced the debut of Gemini 3.7 Flash, while OpenAI introduced the Ultrafast mode for GPT-5.6 Sol. Both advancements are aimed at significantly improving the speed and efficiency of coding tasks and AI agent operations.
Gemini 3.7 Flash Features
Google's Gemini 3.7 Flash is specifically designed for programming and automating business workflows. The model boasts enhancements in multi-step planning, instruction adherence, and code generation capabilities. Importantly, the usage cost has been reduced by 50% compared to its predecessor.
Flash is optimized for scenarios where both speed and output quality are essential, such as in coding and tasks requiring multiple interactions with the AI model.
Previously, Google launched Gemini 3.6 Flash at a rate of $1.5 per million input tokens and $7.5 per million output tokens, reporting a 17% decrease in output token usage compared to the earlier 3.5 Flash model. However, the release of the much-anticipated Gemini 3.5 Pro is still pending, with developers currently testing it with partners.
OpenAI's GPT-5.6 Sol Enhancement
OpenAI's Ultrafast mode for GPT-5.6 Sol promises to deliver performance up to 14 times faster than the standard version, enabling the generation of 750 output tokens per second. This enhancement leverages the Cerebras infrastructure for improved processing speed.
Currently, access to Ultrafast is limited to select clients via API, with plans to expand availability as OpenAI increases its computing resources.
The impressive output rate of 750 tokens per second was previously highlighted during the GPT-5.6 announcement in June, where OpenAI indicated that the use of Cerebras would facilitate this speed, although immediate access for all users was not guaranteed. Additionally, in August, SpaceXAI introduced Grok 4.6, a new model aimed at long-term agent tasks and interactive application development.
