OpenAI GPT-4o Mini Costs Less & Wallops Competition

OpenAI rolled retired GPT-4o mini, a replacement for GPT 3.5 Turbo that is much powerful than different models successful its class. Because it’s hyper efficient, GPT 4o mini will make AI disposable to much group astatine a cheaper value done amended end-user applications.

GPT-4o mini

GPT-4o mini is simply a highly businesslike type of GPT-4o that is cheaper to tally and is fast. Despite it’s nickname arsenic “mini” this connection exemplary is outperforms GPT-4 and GPT-3.5 turbo, arsenic good arsenic solidly outperforming Google’s comparable model, Gemini Flash 1.5.

Preliminary scores by nan unfastened root Large Language Model Systems Organizations shows GPT-4o Mini outperforming Anthropic’s Claude 3 Opus and Google’s Gemini Flash 1.5 and reaching benchmark scores that are comparable to GPT 4.5 Turbo and Gemini 1.5 Pro.

Screenshot Of Language Model Scores

Congrats @openai connected nan caller GPT-4o mini release!

GPT-4o mini's early type "upcoming-gpt-mini" was tested successful Arena successful nan past week.

With complete 6K personification votes, we are excited to stock its early people reaching GPT-4-Turbo performance, while offering important costs reduction… https://t.co/GQZtFVoGXD pic.twitter.com/PAZFjSFVu6

— lmsys.org (@lmsysorg) July 18, 2024

Cost Effective Language Model

An important characteristic of GPT-4o mini is that it’s cheaper to use, 60% cheaper than GPT 3.5 Turbo, which intends that companies that make AI products based connected OpenAI connection models will beryllium capable to connection precocious capacity AI applications that costs importantly less. This makes AI disposable to much group astir nan world.

According to OpenAI:

“Today, we’re announcing GPT-4o mini, our astir cost-efficient mini model. We expect GPT-4o mini will importantly grow nan scope of applications built pinch AI by making intelligence overmuch much affordable. GPT-4o mini scores 82% connected MMLU and presently outperforms GPT-41 connected chat preferences successful LMSYS leaderboard(opens successful a caller window). It is priced astatine 15 cents per cardinal input tokens and 60 cents per cardinal output tokens, an bid of magnitude much affordable than erstwhile frontier models and much than 60% cheaper than GPT-3.5 Turbo.

a matter and imagination exemplary successful nan Assistants API, Chat Completions API, and Batch API. Developers salary 15 cents per 1M input tokens and 60 cents per 1M output tokens (roughly nan balanced of 2500 pages successful a modular book). We scheme to rotation retired fine-tuning for GPT-4o mini successful nan coming days.”

GPT-4o mini Availability

GPT 4o mini is disposable coming to users of ChatGPT Free, Plus and Team, pinch GPT-3.5 nary longer a selectable option. Enterprise users will person entree adjacent week.

Read nan charismatic announcement:

GPT-4o mini: advancing cost-efficient intelligence

Featured Image by Shutterstock/Dean Drobot