• The Input price is what you pay for everything you send to the model.
• The Output price is what you pay for the text the AI sends back to you.
• The “DeepSeek Effect”: DeepSeek R1 is roughly 10x–20x cheaper than GPT or Claude for complex reasoning. Most startups are now using DeepSeek for the “back-end” logic and GPT/Claude for the “front-end” user experience.
• Grok’s Niche: Grok is the most expensive per token, but its value comes from its unique data source. It is the only model that can accurately tell you what happened on the internet 30 seconds ago.
• The Context King: Gemini 3 remains the best value if you have massive files. While its token price is mid-range, its context caching (paying less to “remember” the same 1M token file) makes it the cheapest for long-running research projects.

• The Pro models from GPT, Gemini, Grok and Claude are roughly $20 per month.
• DeepSeek’s Disruption: DeepSeek remains the only major player without a forced monthly subscription. You can use their top-tier model for free on the web, or pay as little as $5 for millions of tokens via their API, making it nearly 95% cheaper than ChatGPT for heavy users.

• Google took the #1 spot on the leaderboard. Users highly rate its “Infinite Memory” (2M+ context window) and its ability to search Google in real-time without hallucinating.
• Grok’s Personality Surge (1,482): Grok 4.1 has become a user favorite for non-corporate tasks. Its “Thinking” mode, combined with an uncensored personality and real-time access to X (formerly Twitter) trends, has pushed it past GPT-5 in general human preference.
• GPT-5.2 Reliability (1,470): While no longer the “Elo King,” ChatGPT remains the most trusted for general tasks. Its user rating is buoyed by its high-quality voice mode and seamless image/video tools (Sora 2).
• The “Claude” Loyalists (1,466): Claude remains the highest-rated model for coding and technical writing. Developers consistently rate it higher than any other model for its human-like nuance and “Agentic” capabilities.
• The “Value King” Deepseek: While DeepSeek has the lowest revenue and a slightly lower Elo score (1,426), it is the most significant market disruptor. Because its API is 94% cheaper than GPT-5, it has become the default “back-end” for thousands of new AI startups in 2026, sacrificing direct revenue for massive market volume.

Context Window