Claude Opus 5 Secures Top Position on Artificial Analysis Intelligence Leaderboard Amidst Fierce Model Competition

Key Takeaways

• Claude Opus 5 Dominance: Claude Opus 5 (max) and Claude Opus 5 (xhigh) are currently ranked as the highest intelligence models globally.
• Speed Leaders: Mercury 2 and HyperNova 60B 2605 lead the industry in output speed, reaching up to 939 tokens per second.
• Massive Context Windows: Llama 4 Scout has set a new benchmark with a 10-million-token context window, significantly ahead of competitors.
• Zero-Cost Entry: New models like Devstral 2 and North Mini Code are entering the market at a $0.00 price point per million tokens.
• Rigorous Methodology: The rankings are based on the Artificial Analysis Intelligence Index v4.1, which utilizes nine specialized evaluation frameworks.

In-Depth Analysis

The Intelligence Hierarchy: Claude vs. GPT

According to the latest rankings from Artificial Analysis, the hierarchy of artificial intelligence has seen a notable shift. Claude Opus 5, specifically the 'max' and 'xhigh' configurations, has ascended to the top of the Intelligence Index. This placement positions Anthropic's flagship model above other high-tier contenders such as Claude Fable 5 (with fallback) and OpenAI's GPT-5.6 Sol (max).

The intelligence rankings are not based on a single metric but are derived from the Artificial Analysis Intelligence Index v4.1. This comprehensive framework incorporates nine distinct evaluations designed to test various facets of machine reasoning and knowledge. These include GDPval-AA v2, τ³-Banking, Terminal-Bench v2.1, SciCode, and the challenging 'Humanity's Last Exam.' Other benchmarks in the suite include GPQA Diamond, CritPt, AA-Omniscience, and AA-LCR. By aggregating these results, the index provides a weighted view of a model's ability to handle complex reasoning, coding, and general knowledge tasks.

Performance Metrics: Speed and Latency Breakthroughs

While intelligence is a primary focus for many users, the Artificial Analysis data highlights that performance in terms of speed and latency is equally competitive. Mercury 2 has emerged as the fastest model in the current lineup, delivering an impressive 939 tokens per second (t/s). It is followed by HyperNova 60B 2605, which clocks in at 436 t/s. Other notable mentions in the speed category include Granite 4.0 H Small and Gemini 3.5 Flash-Lite, which prioritize rapid output for real-time applications.

Latency, or the time it takes for a model to begin responding, is another critical metric where Google's Gemini series shows strength. Gemini 2.5 Flash-Lite boasts the lowest latency at 0.35 seconds, closely followed by Command A+ at 0.41 seconds. Other low-latency performers include Gemini 2.5 Flash and NVIDIA Nemotron 3 Nano. These metrics suggest a bifurcated market where some models are optimized for deep reasoning (like Opus 5) while others are engineered for instantaneous interaction and high-throughput tasks.

Economics and Scale: Pricing and Context Windows

The economic landscape of AI models is becoming increasingly diverse. The data identifies Devstral 2 and North Mini Code as the most cost-effective options, currently listed at $0.00 per million tokens. This zero-cost tier is followed by the Gemma 3 series (4B and 27B variants), which offer competitive pricing for developers looking to scale applications without the high overhead of premium models.

In terms of context window capacity—the amount of data a model can process in a single session—Llama 4 Scout leads the industry with a massive 10-million-token window. This is significantly larger than the 2-million-token window offered by Grok 4.20 0309. Other models with substantial context capabilities include Gemini 1.5 Pro (May version) and Grok 4.1 Fast. The ability to handle 10 million tokens represents a significant leap for the Llama series, allowing for the processing of entire libraries or massive codebases in a single prompt.

Industry Impact

The current rankings on the Artificial Analysis Intelligence Leaderboard signal a period of intense diversification in the AI industry. The fact that Claude Opus 5 has overtaken GPT-5.6 Sol in intelligence metrics suggests that the gap between the top-tier providers is narrowing, with Anthropic currently holding the edge in reasoning capabilities. This competition is likely to drive further innovation in model architecture and training methodologies.

Furthermore, the emergence of zero-cost models and extremely high-speed models like Mercury 2 indicates that the industry is moving toward specialized solutions. Developers no longer have to rely on a single "best" model; instead, they can choose models based on specific needs—whether that is the deep intelligence of Opus 5, the massive context of Llama 4 Scout, or the rapid-fire response of Gemini 2.5 Flash-Lite. The inclusion of complex benchmarks like "Humanity's Last Exam" in the Intelligence Index v4.1 also reflects an industry-wide shift toward more rigorous and human-centric evaluation standards to distinguish between increasingly capable models.

Frequently Asked Questions

Question: Which AI model is currently ranked as the most intelligent?

According to the Artificial Analysis Intelligence Index v4.1, Claude Opus 5 (max) and Claude Opus 5 (xhigh) are the highest-ranked models for intelligence, followed by Claude Fable 5 and GPT-5.6 Sol (max).

Question: What is the fastest AI model in terms of output speed?

Mercury 2 is the fastest model recorded, achieving an output speed of 939 tokens per second. HyperNova 60B 2605 follows it with 436 tokens per second.

Question: Which model offers the largest context window for processing data?

Llama 4 Scout currently offers the largest context window at 10 million tokens, which is substantially larger than the 2-million-token window provided by Grok 4.20 0309.