Google DeepMind, OpenAI Unveil Gemini 3.7 Flash and 750-Token GPT-5.6 Mode
Updated
Updated · Geeky Gadgets · Aug 14
Google DeepMind, OpenAI Unveil Gemini 3.7 Flash and 750-Token GPT-5.6 Mode
3 articles · Updated · Geeky Gadgets · Aug 14
Summary
Gemini 3.7 Flash lifted code-quality success to 43.6% from 34.4%, with Google positioning it as a mid-tier enterprise model for document summarization, video analysis and web development.
Benchmarks highlighted its long-context strengths: 85.4% in long-video understanding versus GPT-5.6 Terra’s 78.9%, and 97% long-context performance against 93.5%.
OpenAI’s GPT-5.6 Ultra-Fast Mode, powered by Cabus chips, reaches 750 output tokens per second—14 times standard speed—for customer support, commerce and financial analysis.
Availability and pricing now diverge: Gemini 3.7 Flash keeps promotional pricing through end-2026 before expected 2027 increases, while GPT-5.6 Ultra-Fast Mode is on a waitlist with pricing still undisclosed.
The launches extend a rapid Gemini 3.7 rollout already underway in GitHub Copilot and Google’s Gemini app, underscoring a broader race to tailor AI for enterprise speed and specialization.