Top AI Models of December 2025: Discover the Complete Performance Rankings!

The December 2025 AI model rankings reveal that Gemini 3 Pro continues to dominate the leaderboard at LMArena. While Google maintains its lead, the real surprises come from Anthropic, which now has three models in the top 10, bolstering its position as a contender against the giants. More notably, OpenAI’s highly anticipated new model, GPT-5.2, fails to make the top 10, as its earlier versions gradually decline. Let’s take a closer look at the top 20 AI models of the month.

Top 10 AI Models of December 2025

Gemini 3 Pro continues its reign over the LMArena. Google’s model outperforms its rivals in creative writing, mathematical reasoning, and prompt adherence. Its newly launched “flash” version has already claimed the third spot.

December brings a fresh highlight with the debut of Claude Opus 4.5 and its “thinking” variant securing the third and fourth spots, respectively. Anthropic thereby maintains three models in the top 10, affirming the rising influence of the Claude series against its competitors. The “thinking” version of Grok 4.1, released by xAI in November 2025, holds onto second place, celebrated for its efficiency in creative, emotional, and collaborative interactions, despite some issues with misinformation and denialism.

Conversely, OpenAI is experiencing some challenges. The major letdown comes from GPT-5.2, the new flagship model of the company, which does not even appear in the top 10 of the general rankings. This notable absence highlights OpenAI’s struggles to keep up with the competition, with only GPT-5.1 “high” making it to the top 10.

Here are the ten most effective models on the market for all tasks combined, as of December 2025:

1. Gemini 3 Pro
2. Grok 4.1 “thinking”
3. Gemini 3 Flash
4. Claude Opus 4.5 “thinking”
5. Claude Opus 4.5
6. Grok 4.1
7. Gemini 3 Flash “thinking”
8. GPT-5.1 “high”
9. Gemini 2.5 Pro
10. Claude Sonnet 4.5 “thinking”

See the full ranking

Best Performing AI Models by Category in December 2025

Beyond its general leaderboard, LMArena also offers rankings by category, providing a more detailed evaluation of each model’s strengths and weaknesses for specific tasks. Here are the top performers by domain in December 2025:

• Text Generation: Gemini 3 Pro remains in first place, followed by Grok 4.1 “thinking” and Claude Opus 4.5 “thinking”. The two Claude models round out the top 5, confirming Anthropic’s ascent in this category.
• Web Development: Claude Opus 4.5 “thinking” takes the lead, with GPT-5.2 “high” making a debut in second place. Claude Opus 4.5 completes the podium, while Google claims the remaining honorable mentions.
• Vision: It’s nearly a clean sweep for Google, which occupies four of the top five spots. Gemini 3 Pro leads, followed by its Flash versions. OpenAI secures a consolation prize with GPT-5.1 “high.”
• Image Generation: A small consolation for OpenAI, as its brand-new image model takes the top spot directly ahead of two versions of Google’s Nano Banana Pro. Two Flux models round out the top five.

How Are AI Models Ranked by LMArena?

LMArena utilizes a participatory methodology to establish an objective ranking of artificial intelligence models. The platform anonymously submits the same request to two different models, and then users vote for the response they find most satisfactory. This direct comparison system generates an Elo score for each model, which adjusts based on the results of each matchup. The key principle: defeating a higher-ranked opponent significantly boosts the score, while losing to a less-ranked model results in a substantial point deduction.