MCP-Universe benchmark exhibits GPT-5 fails greater than half of real-world orchestration duties

The adoption of interoperability requirements, such because the Mannequin Context Protocol (MCP), can present enterprises with insights into how brokers and fashions operate exterior their walled confines. Nonetheless, many benchmarks fail to seize real-life interactions with MCP. 

Salesforce AI Analysis developed a brand new open-source benchmark it calls MCP-Universe, which goals to trace LLMs as these work together with MCP servers in the true world, arguing that it’s going to paint a greater image of real-life and real-time interactions of fashions with instruments enterprises really use. In its preliminary testing, it discovered that fashions like OpenAI’s not too long ago launched GPT-5 are robust, however nonetheless don’t carry out as nicely in real-life eventualities. 

“Present benchmarks predominantly concentrate on remoted features of LLM efficiency, corresponding to instruction following, math reasoning, or operate calling, with out offering a complete evaluation of how fashions work together with real-world MCP servers throughout numerous eventualities,” Salesforce mentioned in a paper. 

MCP-Universe captures mannequin efficiency by software utilization, multi-turn software calls, lengthy context home windows and huge software areas. It’s grounded on present MCP servers with entry to precise knowledge sources and environments. 

Junnan Li, director of AI analysis at Salesforce, instructed VentureBeat that many fashions “nonetheless face limitations that maintain them again on enterprise-grade duties.”

“Two of the largest are: Lengthy context challenges, fashions can lose observe of knowledge or battle to motive persistently when dealing with very lengthy or complicated inputs,” Li mentioned. “And, Unknown software challenges, fashions usually aren’t capable of seamlessly use unfamiliar instruments or techniques in the way in which people can adapt on the fly. Because of this it’s essential to not take a DIY strategy with a single mannequin to energy brokers alone, however as an alternative, to depend on a platform that mixes knowledge context, enhanced reasoning, and belief guardrails to actually meet the wants of enterprise AI.”

MCP-Universe joins different MCP-based proposed benchmarks, corresponding to MCP-Radar from the College of Massachusetts Amherst and Xi’an Jiaotong College, in addition to the Beijing College of Posts and Telecommunications’ MCPWorld. It additionally builds on MCPEvals, which Salesforce launched in July, which focuses primarily on brokers. Li mentioned the largest distinction between MCP-Universe and MCPEvals is that the latter is evaluated with artificial duties. 

The way it works

MCP-Universe evaluates how nicely every mannequin performs a sequence of duties that mimic these undertaken by enterprises. Salesforce mentioned it designed MCP-Universe to embody six core domains utilized by enterprises: location navigation, repository administration, monetary evaluation, 3D design, browser automation and internet search. It accessed 11 MCP servers for a complete of 231 duties. 

• Location navigation focuses on geographic reasoning and the execution of spatial duties. The researchers tapped the Google Maps MCP server for this course of. 
• The repository administration area appears at codebase operations and connects to the GitHub MCP to show model management instruments like repo search, subject monitoring and code modifying. 
• Monetary evaluation connects to the Yahoo Finance MCP server to judge quantitative reasoning and monetary market decision-making.
• 3D design evaluates using computer-aided design instruments by the Blender MCP.
• Browser automation, related to Playwright’s MCP, checks browser interplay.
• The net looking out area employs the Google Search MCP server and the Fetch MCP  to verify “open-domain info in search of” and is structured as a extra open-ended job. 

Salesforce mentioned that it needed to design new MCP duties that replicate actual use circumstances. For every area, they created 4 to 5 sorts of duties that the researchers assume LLMs can simply full. For instance, the researchers assigned the fashions a objective that concerned route planning, figuring out the optimum stops after which finding the vacation spot. 

Every mannequin is evaluated on how they accomplished the duties. Li and his group opted to observe an execution-based analysis paradigm quite than the extra frequent LLM-as-a-judge system. The researchers famous the LLM-as-a-judge paradigm “will not be well-suited for our MCP-Universe situation, since some duties are designed to make use of real-time knowledge, whereas the information of the LLM decide is static.”

Salesforce researchers used three kinds of evaluators: format evaluators to see if the brokers and fashions observe format necessities, static evaluators to evaluate correctness over time and dynamic evaluators for fluctuating solutions like flight costs or GitHub points.

“MCP-Universe focuses on creating difficult real-world duties with execution-based evaluators, which may stress-test the agent in complicated eventualities. Moreover, MCP-Universe gives an extendable framework/codebase for constructing and evaluating brokers,” Li mentioned. 

Even the large fashions have hassle

To check MCP-Universe, Salesforce evaluated a number of widespread proprietary and open-source fashions. These embody Grok-4 from xAI, Anthropic’s Claude-4 Sonnet and Claude 3.7 Sonnet, OpenAI’s GPT-5, o4-mini, o3, GPT-4.1, GPT-4o, GPT-oss, Google’s Gemini 2.5 Professional and Gemini 2.5 Fkash, GLM-4.5 from Zai, Moonshot’s Kimi-K2, Qwen’s Qwen3 Coder and Qwen3-235B-A22B-Instruct-2507 and DeepSeek-V3-0304 from DeepSeek. Every mannequin examined had not less than 120B parameters.

In its testing, Salesforce discovered GPT-5 had one of the best success fee, particularly for monetary evaluation duties. Grok-4 adopted, beating all of the fashions for browser automation, and Claude-4.0 Sonnet rounds out the highest three, though it didn’t submit any efficiency numbers larger than both of the fashions it follows. Amongst open-source fashions, GLM-4.5 carried out one of the best. 

Nonetheless, MCP-Universe confirmed the fashions had issue dealing with lengthy contexts, particularly for location navigation, browser automation and monetary evaluation, with effectivity falling considerably. The second the LLMs encounter unknown instruments, their efficiency additionally drops. The LLMs demonstrated issue in finishing greater than half of the duties that enterprises sometimes carry out.

“These findings spotlight that present frontier LLMs nonetheless fall quick in reliably executing duties throughout numerous real-world MCP duties. Our MCP-Universe benchmark, due to this fact, supplies a difficult and essential testbed for evaluating LLM efficiency in areas underserved by present benchmarks,” the paper mentioned. 

Li instructed VentureBeat that he hopes enterprises will use MCP-Universe to realize a deeper understanding of the place brokers and fashions fail on duties in order that they’ll enhance both their frameworks or the implementation of their MCP instruments.