OpenAI Releases GPT-4o Mini
In a blog post, the corporate introduced the general public launch of its newest AI mannequin. One of the principle highlights of the brand new AI mannequin is its cost-effectiveness. The firm says it’s 60 % cheaper than GPT-3.5 Turbo, which was its earlier most cost-effective small mannequin. In precise numbers, processing a million enter tokens will price the corporate $0.15 (roughly Rs. 12), and per million output tokens will price $0.60 (roughly Rs. 50).
Since GPT-4o Mini is a small mannequin, it additionally provides low latency (time taken to point out the response) which makes it one of many firm’s most effective AI fashions. It at the moment helps textual content and imaginative and prescient (picture processing) within the API, and OpenAI says it should help textual content, picture, video and audio as each enter and output sooner or later.
While the AI agency didn’t share the parameter measurement of the brand new SLM, it revealed that GPT-4o Mini has a context window of 1,28,000 tokens and may help as much as 16,000 output tokens per request (on API). Its data cut-off is as much as October 2023. Coming to efficiency, OpenAI stated the AI mannequin scored 82 % on the Massive Multitask Language Understanding (MMLU) benchmark and outperformed the earliest mannequin of GPT-4 on the unbiased AI chatbot rating web site Large Model Systems Organization’s (LMSYS) leaderboard. However, the adjustments haven’t mirrored on the website on the time of scripting this.
OpenAI additionally claimed that the GPT-4o Mini outperformed Gemini Flash and Claude Haiku in reasoning textual content and vision-based duties, based mostly on its inside testing. It can be claimed to attain 87 % on the Multilingual Grade School Math (MGSM) benchmark and 87.2 % on the HumanEval benchmark. These two benchmarks measure the AI mannequin’s math and coding proficiency, respectively.
For security, the AI agency stated it has used each automated and human evaluations as per its Preparedness Framework. The firm additionally examined the AI mannequin with 70 exterior specialists from completely different disciplines to determine potential dangers.
The GPT-4o Mini has been rolled out because the default AI mannequin for ChatGPT without cost, Plus, and Team customers. The AI mannequin shall be prolonged to enterprise customers beginning subsequent week. Further, it is usually out there as a textual content and imaginative and prescient mannequin within the Assistants API, Chat Completions API, and Batch API. Fine-tuned variants of the mannequin will even be launched sooner or later.
