Google and OpenAI Drop New Models: Speed vs. Soul
Google and OpenAI release new AI models: Gemini 3.1 Flash-Lite for speed and cost efficiency, GPT-5.3 Instant for natural conversation and reduced hallucinations.
“AI Disruption” Publication 9000 Subscriptions 20% Discount Offer Link.
Today, two major tech giants, Google and OpenAI, went head-to-head by successively releasing new versions of their large language models: Gemini 3.1 Flash-Lite from Google and GPT-5.3 Instant from OpenAI.
Google states that Gemini 3.1 Flash-Lite is specifically designed for large-scale intelligence. It is the most cost-effective model in the Gemini 3 series to date, priced at $0.25 per million input tokens and $1.50 per million output tokens. Even at a far lower cost than larger models, it still delivers significantly enhanced performance.
According to benchmarks from Artificial Analysis, while maintaining equivalent or even higher quality, Gemini 3.1 Flash-Lite achieves a Time to First Token (TTFT) that is 2.5 times faster than Gemini 2.5 Flash, with output speed improved by 45%.
GPT-5.3 Instant, on the other hand, has improvements in tone, relevance, and conversational flow, along with a lower refusal rate. Compared to its predecessor, hallucinations are reduced by up to 26.8%, and this model is supported in both ChatGPT and the API.
