Artificial Analysis Intelligence Index: Claude Opus 5 Tops Chart at 63, Anthropic Sweeps Top 10

Claude Opus 5 (max) tops the Artificial Analysis Intelligence Index with a score of 63, narrowly ahead of Claude Fable 5 at 62 — giving Anthropic the top two spots. GPT-5.6 Sol and Grok 4.6 tie for third at 61, and the entire top 10 spans just 8 points, from 63 down to 55.

Key takeaways

• Claude Opus 5 (max) leads the Intelligence Index with a score of 63, the highest of any model listed.
• Anthropic places all three of its models — Claude Opus 5, Claude Fable 5, and Claude Sonnet 5 — inside the top 10.
• GPT-5.6 Sol (max) and Grok 4.6 (high) are tied for third place at 61 points each.
• The spread between rank #1 (63) and rank #10 (55) is only 8 points, one of the tightest top-10 clusters recorded.
• Version upgrades still matter: Grok 4.6 beats Grok 4.5 by 5 points, and GPT-5.6 Sol beats GPT-5.6 Terra by 4 points.

What does the full Intelligence Index ranking look like?

The table below reproduces all ten scores from the Artificial Analysis Intelligence Index as published on August 13, 2026.

Rank	Model	Score
1	Claude Opus 5 (max)	63
2	Claude Fable 5 (with fallback)	62
3	GPT-5.6 Sol (max)	61
3	Grok 4.6 (high)	61
5	Kimi K3 (max)	60
6	Qwen3.8 Max	58
7	Muse Spark 1.2 (xhigh)	57
7	GPT-5.6 Terra (max)	57
9	Grok 4.5 (high)	56
10	Claude Sonnet 5 (max)	55

Why does Anthropic dominate the top 10?

Anthropic is the only company with three separate models in the top 10, occupying ranks 1, 2, and 10. That means every listed Claude variant — Opus 5, Fable 5, and Sonnet 5 — cleared the 55-point threshold, while OpenAI and xAI each contribute two models to the same list.

This breadth matters more than any single score. A company placing one model at the top can happen by chance of tuning for a specific benchmark cycle. Placing three distinct models across the full width of the top 10 — from the leading 63 down to the bottom-anchor 55 — signals that Anthropic's underlying training approach scales consistently across its Opus, Fable, and Sonnet tiers rather than depending on one flagship release.

How tight is the race between GPT-5.6 Sol and Grok 4.6?

GPT-5.6 Sol (max) and Grok 4.6 (high) are exactly tied at 61 points, sharing third place. That tie sits just 2 points behind the Claude Fable 5 score of 62 and only 2 points ahead of Kimi K3 at 60, illustrating how compressed the mid-table has become.

OpenAI and xAI each place two models in the top 10: GPT-5.6 Sol and GPT-5.6 Terra for OpenAI, Grok 4.6 and Grok 4.5 for xAI. The fact that Sol and Terra (57) sit 4 points apart, and Grok 4.6 and Grok 4.5 (56) sit 5 points apart, shows that within-company version upgrades currently produce a larger score jump than the gap between different companies' top models.

What does the Intelligence Index actually measure?

The Intelligence Index is a composite score built from standardized reasoning and problem-solving benchmarks — it is explicitly not a measure of inference speed or operating cost. That distinction matters for anyone comparing models for deployment: a model ranked lower on this index could still be cheaper or faster to run, and the index says nothing about either of those tradeoffs.

Because the index isolates raw reasoning capability, small point differences carry real weight. An 8-point total spread across ten models — the entire published top 10 — means each 1-point gap represents a meaningful fraction of the observed range, not noise. That's why a tied score, like the 61 shared by GPT-5.6 Sol and Grok 4.6, is treated as a genuine dead heat rather than a rounding artifact.

Is version tier upgrading more important than switching providers?

Within this snapshot, upgrading a model's own tier produced bigger score gains than the gap between competing companies' flagship models. Grok 4.6 (high) scores 5 points above Grok 4.5 (high), and GPT-5.6 Sol (max) scores 4 points above GPT-5.6 Terra (max) — both larger than the 1-point gap separating GPT-5.6 Sol from Claude Fable 5.

This pattern suggests that, at least for this index cycle, staying current with a provider's latest tier can matter more than choosing a different provider altogether. The overall competitive picture increasingly reads as tiers-within-companies rather than a clean company-versus-company contest, since the top-10 spread of 8 points is smaller than some single-provider version-to-version jumps.

Bottom line

Anthropic's sweep of ranks 1, 2, and 10 with Claude Opus 5, Fable 5, and Sonnet 5 is the standout structural fact in this Intelligence Index snapshot, while the 8-point total spread across the top 10 confirms how tightly bunched frontier reasoning performance has become. With GPT-5.6 Sol and Grok 4.6 tied at 61, and version-tier gains outpacing company-versus-company gaps, the next index update is worth watching for whether any single model breaks decisively clear of this cluster. For readers tracking how these AI benchmarks intersect with on-chain data and market infrastructure, see the broader research section, and for how this site scores and verifies data points, check the methodology.