LLM war heats up: Claude vs. ChatGPT

AI innovation remains hot and the competition is stiff. OpenAI and Anthropic continue to battle, most recently with the April releases of GPT-5.5 and Claude Opus 4.7. "It's an AI arms race and it's too early to say who wins," Jeff McMillan , founder of McMillanAI and former head of Firmwide AI at Morgan Stanley , told FinAi News . “Every LLM provider will pour resources to build the best model for differentiation in a crowded marketplace.” Each updated model is performing better than the last — in different ways. Read Part 2: Vetting LLMs for banks: ‘No single model can be perfect at everything’ ChatGPT-5.5 leads Opus 4.7 in accuracy, according to Terminal-Bench’s Terminal-Bench 2.0 dataset, an open-source evaluation dataset that measures how well agents operate. However, SWE-bench, a framework that evaluates LLMs on software engineering tasks, says Opus 4.7 takes the cake, for now. Model accuracy

Model Terminal-Bench 2.0 SWE Bench GPT-5.5 82.7% accuracy 58.6% Opus 4.7 6.4% accuracy 64.3% For financial institutions, the procurement question is less settled than the leaderboard suggests, but the scales are increasingly tipping toward Claude. Claude gaining on ChatGPT in adoption The Ramp AI Index's April update showed overall business AI adoption rose to 50.4% of businesses in March compared with 35% a year ago. Anthropic has gained substantial ground in chasing OpenAI, with 30.6% of businesses using Claude in March, up from 24.4% in February. Nearly 35.5% of businesses used OpenAI in March, compared to 34.4% in February, according to the Ramp report. According to the AI index, 51.9% of companies in financial services sector use Anthropic, compared to 45.5% companies using OpenAI. [caption id="attachment_57384711" align="alignnone" width="1563"] (Created by FinAi News with ChatGPT)[/caption] Building for financial services How is Claude gaining on ChatGPT so quickly? Where Anthropic has separated itself from its competitors is in making specific products for finance. Despite financial services being a young vertical at Anthropic, the 10-man team at the company spends hours with its financial services clients to understand bottlenecks and develop tools to remove them, Nicholas Lin , head of product for financial services at Anthropic, told FinAi News . “We are developing a taste for building tools which our financial services clients might use,” Lin said. In the past year, Anthropic has released AI tools built for financial services. In February, the company released: Claude Opus 4.6, which is tuned for financial research that scrutinizes filings, company data and market information; A COBOL conversion tool that helps FIs to upgrade their legacy code infrastructure to new languages; and Customizable plugins for investment banking, financial analysis, private equity and wealth management. "Anthropic has quietly carved out a niche in traditional finance," Arjun Wadwalkar , senior product manager at Global Payments , told FinAi News . "Claude launched specific tools for financial services in the last few months and that is becoming the model of choice on Wall Street." Trust, he argued, is the difference. "Nobody wants to be the first one experimenting in banking,” Wadwalkar said. “As soon as bigger players become associated with Claude, that automatically generates a high level of trust." Goldman Sachs is the marquee case. Chief Information Officer Marco Argenti told CNBC in February that the bank spent six months working with Anthropic engineers to co-develop autonomous agents for trade accounting and client onboarding — applications the firm was "surprised" Claude could handle beyond coding. [caption id="attachment_57384605" align="aligncenter" width="750"] (FinAi analysis)[/caption] Claude better at coding Claude also seems to have the coding edge. "Claude Code is really good. Opus is an incredible model," Ben Conant , chief product officer at Alkami , told FinAi News . But he added a caveat: "We're talking about marginal improvements. Maybe it's 50%, maybe it's 30%. Now, if your job is to write code all day, every day, you really care about 20% or 50%,” Conant said. Megabanks, fintechs and organizations that ship a lot of software will likely pivot toward Claude if marginal gains in coding persist, Conant said, adding that Alkami’s preferred model for quite some time has been Claude but the company is constantly evaluating all options. Spend management provider Extend Chief Executive Officer Andrew Jamison agreed, adding that Claude Code helped the company’s software developers write over 160,000 lines of code, debugging it and get it almost ready to ship within 48 hours. “You do need humans in the last mile, but the speed and quality of Claude Code surprised us,” Jamison said. Extend started using OpenAI’s ChatGPT as its first LLM in 2023 but has been working more on Claude for technical tasks like coding, he said. Multi-model could be winner Having the fastest, cheapest and most accurate model does not guarantee dominance in this AI age, McMillan said, noting that a clear winner might not surface. No matter what model an FI uses, the best technology will be the one that “is tightly tied to your tech stack [and] has been trained to understand your data,” he said. “What FIs need is easy deployment of AI responsibly to drive outcomes.” Moody ’s for one is deploying three different LLMs for its gen AI-driven credit memo tool, Pavle Sabic, senior director of industry practice lead of gen AI at Moody’s, told FinAi News . The tool which can reduce the time needed to generate a credit memo by over 90% uses Google Gemini for web scrapping, Claude to evaluate balance sheets and ChatGPT to generate the report and charts, Sabic said. “Each tool has their own strength and its upto solution providers to blend them to provide the optimal gen AI tool to consumers.” AI business account pricing comparison (as of April) OpenAI (ChatGPT) Anthropic (Claude) Google (Gemini) Business plan price $25/user/month (billed monthly) $20/user/month (billed annually) $25/seat/month (billed monthly) Billed annually for a discount Bundled into Google Workspace Business Standard $14/user/month Business Plus $22/user/month Enterprise plan Custom pricing Custom pricing starts at about $50,000/year for a minimum of 50 seats Custom pricing Despite the ongoing LLM wars, the emerging playbook among major FIs is a multimodel approach, McMillan said. FIs that are moving forward with a multi-model AI approach include: BNY ; Nasdaq ; ING ; JPMorgan ; and Citi . "Financial services companies realize that every model has their own strength and are opting for a multi-model approach rather than getting locked into just one," McMillan said. "It's going to be a similar path to multi-cloud." Register here for the FinAi Lending Summit, set for Oct. 7-8 in Las Vegas.

LLM war heats up: Claude vs. ChatGPT

Part 1: 51.9% of FIs use Claude, compared to 45.5% for OpenAI

Stax chief AI officer breaks down multi-LLM strategy

Citi invests $2B in tech, head of services says

Related Posts

Oracle cut 21,000 jobs in 12 months, says AI replaced some roles

FIs must reconfigure platforms to be AI-discoverable

Robinhood culls 300 in 10% headcount cut as fintechs retrench

Citi invests $2B in tech, head of services says

Stay Informed with Our Newsletters

EMERGING FINTECH DIRECTORY

The Buzz Podcast

SPONSORED

How AI and Product Experts Turn Fuzzy Requirements Into Focused Dev-ready Roadmaps

Is Your Technology Supplier There for You?

Hiding in Plain Sight: How to Use Data to Spot Consumer Accounts Being Used by Small Businesses

Connect

Welcome Back!

Retrieve your password

Continue Reading with FinAi News Premium - Less than $2/Day

LLM war heats up: Claude vs. ChatGPT

Part 1: 51.9% of FIs use Claude, compared to 45.5% for OpenAI

Stax chief AI officer breaks down multi-LLM strategy

Citi invests $2B in tech, head of services says

Related Posts

Stay Informed with Our Newsletters

The Buzz Podcast

Connect

Welcome Back!

Retrieve your password

Unlock This Article

Continue Reading with FinAi News Premium - Less than $2/Day