AI Showdown: GPT-4 vs. Claude 3 Opus – Which Reigns Supreme?

: Amidst the realm of artificial intelligence, 's GPT-4 appears to have slipped into a state of lethargy once more.

chat GPT
chat GPT

However, this time, disheartened patrons of ChatGPT's premium service aren't idly awaiting a swift resolution. Instead, they're turning their gaze towards alternative models, with one particularly piquing their interest: Anthropic's Claude.

The inefficiency of OpenAI's premier model has stirred dissatisfaction among users in recent days. Released in March 2023, GPT-4 users have taken to OpenAI's developer forum and various social platforms to express dismay over its perceived decline in efficacy.

Gripes range from the model's failure to adhere to explicit commands, often providing truncated code instead of complete solutions, to instances where it fails to respond to queries altogether.

“It has become utterly unusable,” lamented one user on OpenAI's online forum last week.

What exacerbates users' frustrations is the recurrence of performance issues in a model heralded as OpenAI's pinnacle offering, for which they pay a monthly fee of $20.

As initially reported by my colleague Alistair Barr, indicators of GPT-4's diminishing capabilities surfaced last summer. The model displayed signs of “weakened logic,” occasionally furnishing incorrect responses to users.

Further evidence of its lethargy emerged earlier this year, prompting OpenAI CEO Sam Altman to publicly acknowledge GPT-4's lackadaisical behavior. He announced on X in February that a remedy had been implemented to address user grievances.

However, when signs of its inefficiency initially surfaced, no other entity had introduced a model that, at least on paper, approached the performance level of GPT-4, thus retaining users' allegiance to the company that arguably sparked the generative craze of the previous year.

Read More: Unlock the Power of GPT-4 Turbo: Microsoft's Game-Changing Upgrade for Copilot Users!

Yet, the landscape has evolved.

With GPT-4 plagued by fresh issues, users have begun exploring alternative models that not only rival OpenAI's flagship offering but arguably surpass it in performance. Enter Anthropic's Claude. This competitor, supported by industry heavyweights like and Amazon, unveiled a premium iteration of its Claude model earlier this month, dubbed Claude 3 Opus—an equivalent to GPT-4.

Upon its release, Anthropic presented data comparing Claude 3 Opus' performance across various benchmarks, such as “undergraduate level knowledge,” “math problem-solving,” “code,” and “mixed evaluations.” Across nearly all metrics, Claude emerged triumphant.

The superiority of Claude isn't merely a claim backed by data. This week, Claude 3 Opus eclipsed GPT-4 on LMSYS Chatbot Arena, an open platform for evaluating AI models.

Naturally, there's a distinction between theoretical prowess and practical application. Consequently, in light of GPT-4's tribulations, even staunch OpenAI adherents have found ample reason to explore alternatives such as Claude.

The satisfaction of many users is palpable.

Following a coding session with Claude 3 Opus, software engineer Anton concluded on X last week that it surpassed GPT-4. “I believe conventional benchmarks fail to capture the true essence of this model,” he remarked.

Angel investor Allie K. Miller noted that GPT-4's performance has noticeably deteriorated in recent months. “The majority of individuals I know are opting for Claude 3,” she observed, alongside Mistral AI's Mixtral 8x7B model.

Wharton professor Ethan Mollick even found Claude 3 to be more adept in J.R.R. Tolkien's constructed Elvish languages of Sindarin and Quenya. “When tasked with translating ‘My hovercraft is full of eels,' Claude 3 provides an original rendition, whereas GPT-4 resorts to web searches,” he disclosed on X.

Meanwhile, on OpenAI's forum, users have lauded Claude for its reliability in coding tasks, likening Claude 3 Opus' performance to the sharpness GPT-4 exhibited upon its debut.

OpenAI declined to comment on GPT-4's performance woes when contacted by BI.

However, some, like Miller, aren't convinced that these issues warrant completely abandoning OpenAI. They speculate that the decline in performance could be attributed to “OpenAI's focus on the next model,” suggesting that resources may be allocated towards that endeavor.

This conjecture may hold weight. As reported by my colleagues Kali Hays and Darius Rafieya earlier this month, OpenAI is poised to unveil GPT-5 by mid-year.

Leave a Reply