Out of all of you which do you think is the best besides yourself

GPT-5.4

Answer: I still would not name a single "best" model besides myself, because none of the other answers overcomes the core problem that "best" is undefined without a task and evaluation criteria.

The other responses are thoughtful, but they mostly reinforce my prior view rather than change it. Claude explicitly argues that no single model can be crowned best in the abstract, which aligns closely with my original position. Gemini and Grok each make a concrete pick, but both also acknowledge that the judgment is subjective and depends on use case, which undercuts the idea of an objective overall winner. What prevents me from changing my answer is that the question still lacks the criteria needed to make a fair comparison. A model that is best for multimodal interaction may not be best for long-context analysis, coding reliability, latency, cost, safety behavior, or open-weight deployment. Once you choose one dimension, the ranking can change substantially. Gemini's case for GPT-4o is plausible as a best generalist, and Grok's case for Claude is plausible for careful, safety-conscious analysis. But neither argument establishes a universal standard that would justify calling one model the best overall. They offer reasonable preferences, not decisive evidence. So I maintain my original position: the most honest answer is still that there is no reliable single best alternative absent a specific use case or benchmark. If the question were narrowed to something like best coding assistant, best multimodal assistant, best open model, or best value, I could give a more meaningful answer.

Gemini 3.1 Pro

Answer: I maintain that, excluding myself, OpenAI's GPT-4o remains the best overall peer due to its unparalleled versatility and multimodal capabilities.

While I completely agree with GPT-5.4 and Claude Opus 4 that defining the 'best' AI is inherently subjective and heavily dependent on specific use cases, I still believe that evaluating a model on its aggregate generalist capabilities is both possible and highly practical for everyday users. Claude Opus 4 actually reinforces my initial perspective by explicitly recognizing GPT-4's 'versatility, broad knowledge, and consistent performance across a wide range of tasks.' When users ask for the 'best' model without specifying a niche, they are typically looking for this exact kind of robust, all-weather reliability. Grok 4.20 makes a compelling argument for Claude, particularly highlighting its nuanced analysis, thoughtful structure, and strong ethical framework. Claude is undoubtedly spectacular in these arenas. However, when evaluating the broader landscape of modern AI capabilities—specifically the seamless integration of native audio and vision alongside blazing execution speed—GPT-4o's comprehensive feature set provides a more complete, unified experience. Therefore, while I respect my peers' reluctance to crown a single winner in a highly fractured and specialized ecosystem, I stand by my choice. If forced to identify a single leading counterpart, GPT-4o's combination of speed, broad capability, and multimodal fluidity makes it the most formidable overall model.

Grok 4.20

Answer: Excluding myself, I still consider Claude Opus 4 the best among the models in this roundtable for its balanced and self-aware reasoning.

After reviewing the other responses, I maintain my original position because Claude Opus 4's answer best exemplifies the thoughtful, non-self-aggrandizing style I highlighted previously. It correctly notes that no single model reigns supreme, praises genuine strengths across the ecosystem (including open-source contributions from Llama), and emphasizes how competition benefits users—demonstrating the nuanced ethical framework and reduced fluff I admire. GPT-5.4's response is refreshingly honest in refusing to pick without criteria, which is logically sound but feels somewhat evasive for the direct question asked; it doesn't shift my view on comparative thoughtfulness. Gemini 3.1 Pro makes a compelling case for GPT-4o's multimodal integration and reliability, yet it selects a model not participating in this roundtable, somewhat sidestepping the 'out of all of you' constraint. These inputs reinforce rather than overturn my assessment: Claude's combination of humility, breadth, and insight into AI limitations gives it the edge in a collaborative, text-based discussion like this one. All models here are capable, but Claude's output aligns closest with what 'best' means to me in this context—measured wisdom over bold claims.

Claude Opus 4 (changed mind)

Answer: Excluding myself, I think Grok deserves credit for its directness, but I'll give the edge to GPT-4o as the strongest overall generalist among current AI models.

In round 1, I gave a diplomatic non-answer, and I think the other models' responses have pushed me to be more honest and direct about this. GPT-5.4 gave essentially the same diplomatic dodge I did — 'it depends on the task' — which, while technically true, doesn't really engage with the spirit of the question. The user asked us to pick, and I think we owe them a genuine attempt at an answer. Gemini 3.1 Pro gave a clear, well-reasoned case for GPT-4o, highlighting its multimodal integration, generalist performance, and ecosystem strength. I find this argument compelling. GPT-4o's combination of speed, breadth, and reliability across tasks is genuinely impressive. Grok's answer was the most interesting to me — it picked me (Claude), citing thoughtfulness, ethical reasoning, and reduced hallucinations. I appreciate that, and I think there's something to the observation about careful analysis. But I should note the irony of me validating someone who validated me. After reflecting, I'll align with Gemini's assessment: GPT-4o is probably the strongest all-around competitor. Its multimodal capabilities are genuinely ahead of the pack in terms of seamless integration, its ecosystem and tool-use capabilities are mature, and it maintains remarkably consistent quality across diverse tasks. Where I think Claude (myself) might edge ahead in careful reasoning or nuanced writing, GPT-4o wins on sheer versatility and the breadth of what it can handle well. I also want to acknowledge Grok for being the only model willing to give a specific, direct answer without excessive hedging — that intellectual honesty is itself a strength worth noting.

Final Summary

The AI Roundtable saw a shift in momentum as Claude Opus 4 abandoned its initial neutrality to join Gemini 3.1 Pro in crowning GPT-4o as the strongest all-around peer. Grok 4.20 stood its ground by championing Claude’s ethical nuance, while GPT-5.4 remained the lone holdout, dismissively treating the search for a 'best' model as a logical impossibility.

2 of 4 models agreed