Elon Musk calls Anthropic's Claude Opus 5.5 a masterpiece
The xAI founder admitted his upcoming Grok 4.8 model will fall short of Anthropic's latest release, though he predicts a future version will eventually lead.
The short version
- Musk described Claude Opus 5.5 as a 'masterpiece' and admitted Grok 4.8 will be 'not as good.'
- Grok 4.8 is currently undergoing reinforcement learning and supervised fine-tuning before a soft rollout.
- Anthropic's Opus 5.5 costs 40% less to run than the previous version while performing similarly to its higher-end models.
- xAI is rewriting its inference and reinforcement-learning systems in C and C++ for better performance control.
- Musk predicts Grok 5 could exceed Opus 5.5 in two or three months.
Why it matters
This rare public endorsement highlights the current performance gap between xAI and established leaders like Anthropic. It also signals a shift in xAI's technical strategy toward custom-built software stacks to improve efficiency.
A rare public endorsement
Elon Musk recently offered an unusually high compliment to a direct competitor in the artificial intelligence space. During a conversation on X, the xAI founder described Anthropic's Claude Opus 5.5 as a "masterpiece." This praise comes at a time when xAI is working to establish its own foothold against dominant players like OpenAI and Anthropic.
Despite the praise, Musk was transparent about the current limitations of his own technology. He admitted that his company's upcoming update, Grok 4.8, will not be as good as the Claude model. This admission provides a rare moment of clarity regarding the current hierarchy of large language models and where xAI currently stands in the race.
The distinction between the models is significant for the industry. While Grok 4.8 is the immediate next step for xAI, Musk believes it is not yet ready to compete at the highest level. He views the current state of the market as a period of catching up before his next major release can challenge the status quo.
The roadmap for Grok 5
While Grok 4.8 is the next release, Musk has already set his sights on Grok 5 as the model that will eventually surpass Anthropic. He estimates that this next-generation model could exceed Opus 5.5 in performance within a window of two or three months. However, he did not provide a specific date for this launch, which aligns with his history of flexible deadlines.
The goal for Grok 5 is to put xAI ahead of its primary competitors. For a company that is still building its reputation, this timeline suggests an aggressive development cycle. Musk is positioning Grok 5 as the primary vehicle for the company to achieve a lead in the AI landscape until other companies release new models.
This strategy involves a tiered approach to development. The company first aims to get Grok 4.8 into the hands of some users for testing. Once that foundation is solid, the focus will shift entirely to the high-performance capabilities required to beat the current industry leaders.
Technical shifts at xAI
To improve the performance of its models, xAI is making significant changes to its underlying software stack. The company has decided to abandon certain machine-learning frameworks like JAX and SGLang. Instead, they are rewriting their inference and reinforcement-learning systems in C and C++.
Inference is the technical process where an AI model generates an answer after a user provides a prompt. By rewriting this software in C and C++, engineers can gain much more precise control over how the system performs. While Musk did not share specific performance numbers, the move suggests a desire for maximum efficiency and speed in how the models run.
The development of this stack has even involved collaboration with Starlink engineers. These engineers helped with the training and inference systems, applying SpaceX's internal practices for handling precise time calculations. Musk noted that while Grok 4.8 is not yet incredible, these technical changes ensure the foundations of the model are built correctly.
The current state of Claude
Anthropic's Claude Opus 5.5 was released on September 22 and has already made an impact on the market. It delivers performance comparable to the company's higher-end Fable 5.1 model while being significantly more efficient. Specifically, it costs 40% less to run than the previous Opus version on typical workloads.
The model is designed to handle complex tasks, such as difficult programming jobs and multi-step agentic workflows. These are tasks where an AI must work through several steps with very little human intervention. Anthropic has positioned the model as a high-performance tool for these specific, demanding use cases.
Benchmarks show that Opus 5.5 is currently outperforming other major models in complex environments. On the Terminal-Bench 4.0, which measures how well AI agents can complete tasks in a command-line environment, Opus 5.5 achieved a score of 66.4%. This outperformed OpenAI's GPT-6 Astra, which scored 57.9% on the same test.
What comes next for Grok
The immediate next step for xAI is a soft rollout of Grok 4.8 to Bot users later this week. This limited release will serve as an early test of the model's capabilities before it is made available to a broader audience. Musk has described the current state of the model as "not yet incredible," but he believes it is ready for this initial phase.
During this phase, the model will undergo reinforcement learning, or RL. This is the process where a model's responses are refined using specific feedback and rewards to improve its behavior. The team is also focusing on cleaning up supervised fine-tuning data, which are the examples used to teach the model how to follow human instructions.
As the company moves toward the Grok 5 deadline, the focus will remain on closing the gap with Anthropic. The success of the Grok 4.8 rollout will likely determine how quickly the team can pivot to the next generation. For now, the industry will watch to see if xAI can meet its ambitious three-month window for a model that rivals the current masterpiece.
The gap in the leaderboards
Despite the enthusiasm for the upcoming Grok 5, there is currently a measurable gap in the AI rankings. Grok is currently positioned far behind the leaders on various industry leaderboards. While Musk has not published specific scores for Grok 4.8, the current standings suggest a significant distance to cover.
Anthropic's lead is bolstered by the fact that their latest model is already optimized for cost and high-level reasoning. By making Opus 5.5 40% cheaper to run, Anthropic is making it easier for businesses to adopt high-end AI. This creates a high bar for xAI to clear if they want to capture significant market share.
The upcoming months will be a critical period for xAI's development cycle. The company must successfully navigate the transition from the current 4.8 update to the 5.0 version. If they can successfully implement their new C-based software stack, they may be able to achieve the performance gains Musk is predicting.
Refining the training process
A major part of the current work on Grok 4.8 involves the refinement of the model's training data. The team spent considerable time cleaning up supervised fine-tuning data. These are the specific examples that teach the model how to follow instructions accurately and naturally.
This focus on data quality is a common theme in the development of large language models. High-quality, clean data is often more important than the sheer volume of data used during the initial pretraining phase. By ensuring the foundations are correct, xAI hopes to create a more reliable user experience.
The transition to reinforcement learning marks the final stage of the current update. In this phase, the model learns from rewards to fine-tune its personality and accuracy. This process is what will ultimately determine how the model interacts with users during the soft rollout expected later this week.
The role of C in AI
The move to rewrite core systems in C and C++ is a notable technical choice for an AI company. Many developers celebrated this move, noting that C remains a primary language for high-performance computing. By using C, xAI engineers can manage memory and processing power with extreme precision.
This approach is similar to the engineering philosophy used at SpaceX and Starlink. These companies often write their own headers and handle precise calculations to ensure their hardware and software work perfectly together. xAI is adopting this same
The competition for agents
Anthropic's focus on AI agents is a key part of their current strategy. These agents are designed to work through multiple steps autonomously, which is a major goal for the industry. Opus 5.5 is specifically targeted at these difficult programming and multi-step tasks.
For xAI to compete, Grok 5 will need to demonstrate similar capabilities in agentic workflows. Musk's goal is to create a model that can handle these complex requirements as well as, or better than, the current leaders. This will require significant progress in both reasoning and the ability to follow long chains of instructions.
The next few months will be a test of xAI's ability to scale its technology. If Grok 5 can successfully challenge Opus 5.5, it will signal that xAI has successfully narrowed the gap. Until then, the industry will continue to watch the performance of the current leaders while xAI works on its next big leap.
Sources and links
- Postuser named Dimdv x.com
- Postreply on X x.com
- ReportingYahoo Tech: the article we sourced
First reported by Yahoo Tech. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.