OpenAI Introduces GPT-4o Mini: Revolutionizing Affordable AI
OpenAI is dedicated to making AI intelligence broadly accessible to everyone. Today, we're excited to announce GPT-4o Mini, our most cost-efficient small model yet. With GPT-4o Mini, we aim to significantly expand the range of AI applications by making intelligence more affordable. This model scores an impressive 82% on MMLU and outperforms GPT-4 on chat preferences in the LMSYS leaderboard. Priced at just 15 cents per million input tokens and 60 cents per million output tokens, GPT-4o Mini is an order of magnitude more affordable than previous frontier models and over 60% cheaper than GPT-3.5 Turbo.
GPT-4o Mini is designed to enable a broad range of tasks with its low cost and low latency. It is ideal for applications that require chaining or parallelizing multiple model calls (such as calling multiple APIs), passing a large volume of context to the model (like a full code base or conversation history), or providing fast, real-time text responses (such as customer support chatbots).
Currently, GPT-4o Mini supports text and vision in the API, with future updates planned to include text, image, video, and audio inputs and outputs. The model boasts a context window of 128K tokens, supports up to 16K output tokens per request, and has knowledge up to October 2023. With an improved tokenizer shared with GPT-4o, handling non-English text is now even more cost-effective.
By introducing GPT-4o Mini, OpenAI is paving the way for more accessible and affordable AI solutions. This groundbreaking model is set to revolutionize the AI landscape, enabling developers and businesses to build innovative applications without breaking the bank.
Stay tuned for more updates on how GPT-4o Mini is making AI intelligence more accessible and affordable than ever before.
GPT-4o Mini: Superior Textual Intelligence and Multimodal Reasoning
GPT-4o Mini is a small model that excels in both textual intelligence and multimodal reasoning, outperforming GPT-3.5 Turbo and other small models on key academic benchmarks. Supporting the same range of languages as GPT-4o, GPT-4o Mini demonstrates impressive performance in function calling, enabling developers to build applications that fetch data or interact with external systems. It also boasts improved long-context performance compared to GPT-3.5 Turbo.
Benchmark Performance:
Reasoning Tasks: GPT-4o Mini surpasses other small models in reasoning tasks involving text and vision, scoring 82.0% on the MMLU benchmark, a measure of textual intelligence and reasoning. This is higher than the 77.9% scored by Gemini Flash and the 73.8% by Claude Haiku.
Math and Coding Proficiency: GPT-4o Mini excels in mathematical reasoning and coding tasks. On the MGSM benchmark, which measures math reasoning, GPT-4o Mini scored 87.0%, outperforming Gemini Flash's 75.5% and Claude Haiku's 71.7%. For coding performance, GPT-4o Mini scored 87.2% on the HumanEval benchmark, compared to Gemini Flash's 71.5% and Claude Haiku's 75.9%.
Multimodal Reasoning: GPT-4o Mini also shows strong performance in multimodal reasoning, scoring 59.4% on the MMMU benchmark. This surpasses Gemini Flash's 56.1% and Claude Haiku's 50.2%.
By delivering superior textual intelligence and multimodal reasoning capabilities, GPT-4o Mini sets a new standard for small AI models. This enhanced performance across various benchmarks makes GPT-4o Mini a powerful tool for developers aiming to create innovative applications.
Stay tuned to our blog for more updates on how GPT-4o Mini is leading the way in affordable, high-performance AI solutions.
Built-in Safety Measures: Ensuring Reliability and Security with GPT-4o Mini
At OpenAI, safety is an integral part of our AI development process, reinforced at every stage. From the outset, we embed safety measures in our models and enhance them continually. During pre-training, we filter out unwanted content, including hate speech, adult content, personal information aggregators, and spam. Post-training, we align the model’s behavior with our policies using techniques like reinforcement learning with human feedback (RLHF) to ensure the accuracy and reliability of our models’ responses.
GPT-4o Mini's Safety Measures:
Comprehensive Safety Mitigations: GPT-4o Mini incorporates the same robust safety mitigations as GPT-4o, thoroughly assessed using both automated and human evaluations under our Preparedness Framework. This aligns with our voluntary commitments to safety. Over 70 external experts from fields such as social psychology and misinformation rigorously tested GPT-4o to identify and address potential risks. We will share these findings in the forthcoming GPT-4o system card and Preparedness scorecard. Insights from these evaluations have significantly enhanced the safety of both GPT-4o and GPT-4o Mini.
Advanced Safety Techniques:
Building on these insights, our teams have employed new techniques to further improve the safety of GPT-4o Mini. The API version of GPT-4o Mini is the first model to implement our instruction hierarchy method, which enhances the model’s ability to resist jailbreaks, prompt injections, and system prompt extractions. This innovation makes the model’s responses more reliable and safer for large-scale application use.
Commitment to Safe AI:
Our dedication to safe AI development ensures that GPT-4o Mini can be used confidently across a variety of applications. By integrating these advanced safety measures, we aim to provide a reliable and secure AI tool that meets the highest standards of accuracy and ethical use.
GPT-4o Mini: Availability and Pricing for Developers and Users
Introducing GPT-4o Mini: Now Available Across APIs and ChatGPT
We are excited to announce that GPT-4o Mini is now available as a text and vision model in the Assistants API, Chat Completions API, and Batch API. Developers can access this advanced AI model at an affordable rate of just 15 cents per 1M input tokens and 60 cents per 1M output tokens, roughly equivalent to 2,500 pages in a standard book. Additionally, we will be rolling out fine-tuning capabilities for GPT-4o Mini in the coming days.
Accessibility for All Users:
Starting today, GPT-4o Mini will be accessible to Free, Plus, and Team users on ChatGPT, replacing GPT-3.5. Enterprise users will gain access starting next week, reinforcing our commitment to making AI benefits widely available.
Affordable AI for Developers:
With competitive pricing and broad availability, GPT-4o Mini empowers developers to create innovative applications without breaking the bank. Its low cost and high performance make it an ideal choice for a wide range of projects, from small-scale applications to large enterprise solutions.
Stay tuned to our blog for more updates on GPT-4o Mini and discover how this cost-efficient, high-performing AI model can transform your projects. By making advanced AI accessible to all, we continue to drive innovation and expand the possibilities of what AI can achieve.
For AI News Join Duniyadari
