Mark Zuckerberg's Meta Says Llama 3 Beats Google's Gemini, Mistral And Jeff Bezos-backed Anthropic's Claude 3, But OpenAI's GPT-4 Is Notably Missing From Its Comparison
Mark Zuckerberg's Meta Says Llama 3 Beats Google's Gemini, Mistral And Jeff Bezos-backed Anthropic's Claude 3, But OpenAI's GPT-4 Is Notably Missing From Its Comparison
Mark Zuckerberg's Meta Platforms Inc. (NASDAQ:META) revealed that its new large language model, Llama 3, has surpassed other AI models in benchmark tests. However, OpenAI's latest flagship model, GPT-4, is notably missing from its comparison.
馬克·扎克伯格的元平台公司（納斯達克股票代碼：META）透露，其新的大型語言模型Llama 3在基準測試中已經超過了其他人工智能模型。但是，OpenAI 的最新旗艦機型 GPT-4 在比較中明顯缺失。
What Happened: The Llama 3 model, which Meta unveiled on Thursday, has outperformed several other AI models in benchmark tests, including those from Alphabet Inc.'s (NASDAQ:GOOGL) (NASDAQ:GOOG) Google, Jeff Bezos-backed Anthropic, and Mistral AI.
發生了什麼：Meta週四發佈的Llama 3模型在基準測試中的表現優於其他幾種人工智能模型，包括來自Alphabet Inc的模型。”s（納斯達克股票代碼：GOOG）（納斯達克股票代碼：GOOG）谷歌、傑夫·貝佐斯支持的Anthropic和Mistral AI。
The new model is expected to be available to cloud providers like Amazon.com Inc.'s (NASDAQ:AMZN) Amazon Web Services (AWS) and model libraries such as Hugging Face soon.
預計新模式將提供給像亞馬遜公司這樣的雲提供商。”s（納斯達克股票代碼：AMZN）亞馬遜網絡服務（AWS）以及諸如Hugging Face之類的模型庫即將推出。
Subscribe to the Benzinga Tech Trends newsletter to get all the latest tech developments delivered to your inbox.
訂閱 Benzinga 技術趨勢時事通訊，將所有最新的技術發展發送到您的收件箱。
The Llama 3 model, which comes in two sizes with 8B and 70B parameters, has shown significant improvements in text-based responses. It has demonstrated a higher diversity in answering prompts, fewer false refusals, and improved reasoning abilities.
Llama 3模型有兩種尺寸，參數爲8B和70B，在基於文本的響應方面已顯示出顯著的改進。它顯示出回答提示的多樣性更高，錯誤的拒絕更少，推理能力的提高。
The model also shows a better understanding of instructions and improved code-writing capabilities.
該模型還顯示出對指令的更好理解和更高的代碼編寫能力。
Meta's blog post stated that Llama 3 outperformed similarly sized models like Google's Gemma and Gemini, Anthropic's Claude 3, and Mistral 7B in specific benchmarking tests.
Meta 的博客文章指出，在特定的基準測試中，Llama 3 的表現優於谷歌的 Gemma 和 Gemini、Anthropic 的 Claude 3 和 Mistral 7B 等類似大小的模型。
In the MMLU benchmark, which measures general knowledge, the 8B version of Llama 3 outperformed both Gemma 7B and Mistral 7B, while the 70B version slightly edged Gemini Pro 1.5.
在衡量常識的MMLU基準測試中，Llama 3的8B版本的表現超過了Gemma 7B和Mistral 7B，而70B版本的表現略高於Gemini Pro 1.5。
It is worth noting that Meta did not mention OpenAI's GPT-4 in its post. The company also highlighted that benchmark testing AI models, while helpful, is not perfect, as the datasets used for benchmarking are often part of the model's training.
值得注意的是，Meta 在其帖子中沒有提到 OpenAI 的 GPT-4。該公司還強調，基準測試人工智能模型雖然有用，但並不完美，因爲用於基準測試的數據集通常是模型訓練的一部分。
Why It Matters: Earlier in March, Anthropic introduced its latest generative AI model, Claude 3, claiming it outperforms rivals like OpenAI's GPT-4.
它爲何重要：3月初，Anthropic推出了其最新的生成式人工智能模型Claude 3，聲稱其表現優於OpenAI的 GPT-4 等競爭對手。
In a similar vein, Elon Musk's xAI announced a major update to Grok with improvements across the board in various metrics.
同樣，埃隆·馬斯克的xAi宣佈對Grok進行重大更新，對各種指標進行了全面改進。
OpenAI, on the other hand, is expected to release a "significantly better" GPT-5 model later this year.
另一方面，預計OpenAI將在今年晚些時候發佈一款 “好得多” 的 GPT-5 機型。
Check out more of Benzinga's Consumer Tech coverage by following this link.
通過以下方式查看 Benzinga 對消費科技的更多報道 點擊這個鏈接。
Disclaimer: This content was partially produced with the help of Benzinga Neuro and was reviewed and published by Benzinga editors.
免責聲明： 該內容部分是在 Benzinga Neuro 的幫助下製作的，並由 Benzinga 編輯審查和出版。
Photo courtesy: Shutterstock
照片來源：Shutterstock
譯文內容由第三人軟體翻譯。