Top 20 AI Models of February 2026: See the Full Performance Rankings!

Last month, we noticed a significant drop in the ranking of ChatGPT at LMArena, which evaluates language models throughout 2025. Until November 2024, OpenAI’s models were the undisputed leaders but began losing ground, plummeting in the rankings from July 2025. Until now, Google seemed to be benefiting from this downturn. However, in February, the company was overtaken by an emerging competitor: Anthropic.

The Top 10 AI Models in January 2026

While Gemini models are known for their consistency, those from Anthropic have shown more volatility. However, since the format change at LMArena in June, the developers of Claude have been steadily climbing the leaderboard. In February 2026, the Opus 4.6 models (and their Thinking version), launched at the start of the month, have taken the top spot from Gemini 3 Pro. They excel in almost all generalist domains, such as responding to lengthy queries and complex prompts, creative writing, and following instructions. The rest of the top 10 includes strong performers like Gemini, ranking solid third, as well as Grok and Dola.

Consequently, OpenAI’s models have been left out of the top 10 for the second consecutive month. This list also no longer includes Ernie, the Chinese model developed by Baidu, confirming the total dominance of U.S. companies in the industry.

Here are the ten most efficient models in the market, across all tasks, as of February 2026:

1. Claude Opus 4.6 “thinking”
2. Claude Opus 4.6
3. Gemini 3 Pro
4. Grok 4.1 “thinking”
5. Gemini 3 Flash
6. Dola Seed 2.0 Pro
7. Claude Opus 4.5 “thinking”
8. Claude Opus 4.5
9. Grok 4.1
10. Gemini 3 Flash “thinking”

Top AI Models by Category

Beyond the general rankings, LMArena offers thematic leaderboards that help pinpoint the strengths of each model according to specific uses. Here are some trends observed in the latest charts:

• Text Generation: Claude Opus 4.6 clearly dominates the category, with its two versions (thinking and standard) taking the top two spots. Gemini 3 Pro maintains a solid position on the podium, while Grok 4.1 “thinking” and Gemini 3 Flash remain close behind.
• Web Development: Claude confirms its leadership in coding. Both versions of Opus 4.6 also lead in this area, followed by Opus 4.5. OpenAI manages to secure the fourth spot with GPT-5.2 “high”.
• Vision: Google leads in image analysis with Gemini 3 (Pro and Flash). Dola Seed 2.0, a Californian company, ranks third. OpenAI, once a pioneer in this field, settles for fifth place.
• Image Generation: a small consolation for OpenAI, which still leads in image generation and editing. However, Gemini remains close behind, followed by Grok.
• Web Search: Google takes the top two spots with Gemini 3 Flash Grounding and Gemini 3 Pro Grounding. OpenAI is close behind with GPT-5.2 Search and GPT-5.1 Search. Meanwhile, Grok and Claude appear lower in the rankings.

How are AI Models Ranked by LMArena?

To provide as impartial a view as possible of the sector, LMArena relies on collective evaluation. The platform organizes anonymous head-to-head competitions between models, all responding to the same prompt, allowing users to directly compare responses and choose the one they deem most complete. Following each matchup, an Elo score is adjusted for each participant. This mechanism rewards victories against higher-ranked models and imposes harsher penalties for losses to less well-positioned competitors, gradually refining the hierarchy.