when grok 3 released it was at the top of lmarena for a while. what was the landscape of…
- User
when grok 3 released it was at the top of lmarena for a while. what was the landscape of other ais at that time, and how long until something replaced it at the top and what was it?
Assistant
Grok-3 (specifically Grok-1, released in November 2023) did indeed briefly top the LMSys Chatbot Arena leaderboard around December 2023 - January 2024. Here's the landscape and what happened next:
Key Competitors at Grok-1's Peak (Late 2023/Early 2024)
- OpenAI's GPT-4 Turbo (Nov 2023):
The dominant leader before Grok-1's rise. Slightly outperformed in coding but trailed Grok-1 in overall Arena Elo. - Anthropic's Claude 2.1 (Nov 2023):
Competitive in reasoning and long-context tasks, but ranked below Grok-1/GPT-4 Turbo. - Google's Gemini Pro (Dec 2023):
Initial release underperformed; ranked below top models. - Open-Source Models (Mixtral, Yi, Llama 2):
Mixtral-8x7B was the strongest open-source model but still trailed top proprietary models by ~100 Elo points.
How Long Did Grok-1 Stay at #1?
- Approximately 6–8 weeks (mid-Dec 2023 to early Feb 2024).
- Grok-1's lead was narrow. By late January 2024, GPT-4 Turbo and Claude 2.1 were statistically tied with it.
What Dethroned Grok-1?
- Claude 3 Opus (released March 4, 2024) decisively took #1:
- Surpassed Grok-1 by ~50 Elo points immediately.
- Dominated in reasoning, math, coding, and long-context tasks.
- Remained #1 for months until GPT-4o arrived.
Timeline Summary
Date Event Leaderboard Impact Nov 2023 Grok-1 (Grok-3) released Reached #1 by mid-Dec 2023 Dec 2023–Jan 2024 GPT-4 Turbo/Claude 2.1 updates Grok-1 tied with rivals Mar 4, 2024 Claude 3 Opus released #1 spot taken, Grok-1 fell to #3 May 2024 GPT-4o & Gemini 1.5 Pro released Grok-1 dropped further (top 10) Why Grok-1's Lead Was Short
- Rapid Model Releases: Claude 3 and GPT-4o leapfrogged capabilities within months.
- Narrow Margins: Grok-1's lead was small (<20 Elo over GPT-4 Turbo).
- Specialized Strengths: Grok-1 excelled in "fun" interactions (aligned with X's platform) but lagged in technical tasks vs. Claude 3/OpenAI.
As of mid-2024, Claude 3 Opus, GPT-4o, and Gemini 1.5 Pro dominate the top, while Grok-1 fell to #8–#10 (behind models like Claude 3 Sonnet and Mixtral 8x22B). Grok's team is developing Grok-1.5 with improved reasoning, but it hasn't challenged the top tier yet.
The AI race remains volatile—expect reshuffling with Grok-2, Claude 4, or Gemini 2.0! 🚀
- OpenAI's GPT-4 Turbo (Nov 2023):
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.