Generative AI Experts: ChatGPT, Gemini, Claude & More

OpenAI CEO Sam Altman has hinted at the upcoming release of GPT-6, the next iteration of the company’s flagship large language model, suggesting it will feature “extra goblins.” While details remain scarce, the comment, made during a recent appearance, signals a continued push for more sophisticated and potentially unpredictable AI capabilities.

Altman’s remark was reported by multiple outlets following his appearance at an event on April 30, 2026. The specific meaning of “extra goblins” remains unclear, but it has sparked speculation within the AI community. Some interpret it as a reference to increased complexity and emergent behaviors within the model, potentially leading to more creative, but also less predictable, outputs.

GPT-5.2 and the Current AI Landscape

The announcement comes as OpenAI continues to refine and deploy GPT-5.2 and GPT-5.3, the current leading models. These models represent a significant leap forward in reasoning, coding, and writing capabilities compared to previous iterations. According to a recent guide published by Ethan Mollick, GPT-5.2/5.3 are currently considered among the most powerful AI models available, alongside Anthropic’s Claude Opus 4.6 and Google’s Gemini 3 Pro.

Mollick’s analysis, published on February 18, 2026, highlights a shift in how AI is being utilized. The focus is moving beyond simple chatbot interactions towards using AI as an “agent” capable of completing tasks autonomously. This requires models with strong reasoning skills and the ability to leverage external tools.

The Competitive AI Market

The generative AI market is intensely competitive, with several key players vying for dominance. Alongside OpenAI, Anthropic (creator of Claude) and Google (with Gemini) are major contenders. Other models, including DeepSeek, Perplexity, Grok, and Qwen, are also making significant strides, each with unique strengths. A report from Laksn.com, published in 2025, details the architectures, performance, pricing, and ideal use cases for these various models.

Claude is particularly noted for its powerful coding abilities and structured responses, while Grok is designed for deep reasoning and real-time analysis. Perplexity excels in fact-based research, integrating real-time search capabilities. Gemini is strong in multimodal analysis, handling images, video, and documents effectively.

Pricing and Access in 2025

Pricing for these models varies considerably. According to a snapshot from December 11, 2025, as reported by SparkAgentAI, ChatGPT (GPT-4o family) ranges from $3 to $15 per 1 million tokens. Claude (3.7 Sonnet / Haiku) is priced between $1 and $15 per 1 million tokens, with Haiku being the most affordable option. Gemini (2.0 Flash / Pro) offers the lowest cost at $0.50 to $10 per 1 million tokens, with Flash being the cheapest LLM on the market.

Independent pricing audits conducted in 2025 indicate that OpenAI leads in agentic cost efficiency, Claude excels in long-context cost predictability, and Gemini offers the most affordable multimodal inference.

The Rise of Agentic AI

The shift towards “agentic AI,” as described by Mollick, is transforming how these models are used. Rather than simply responding to prompts, AI agents can now be assigned tasks and utilize tools to achieve specific goals. This requires models capable of complex reasoning, planning, and execution.

The capabilities of these models are impacting a wide range of industries, from automated content creation and software development to research analysis and customer support. The choice of which model to use is increasingly a strategic business decision, as highlighted by a founder quoted in the SparkAgentAI report: “Claude writes like my smartest teammate, ChatGPT thinks like my sharpest engineer, and Gemini sees the world like my research intern.”

While the meaning of “extra goblins” remains a mystery, Altman’s comment suggests that OpenAI is continuing to push the boundaries of AI capabilities, potentially introducing new levels of complexity and unpredictability with GPT-6.