GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 586 -0.5%
INKL 531 —
CL-OP46 497 —
CL-OP48 490 -0.2%
GLM-52 897 —
GPT-56SC 873 —
CL-OP5X 865 -0.9%
GROK-46H 865 -0.9%
GEM-37FH 865 -0.9%
GPT-56T 861 —
GLM-5 856 —
MUSE-SPK 841 —
QWEN-38X 824 -2.3%
GPT-6A 820 —
KIMI-K3X 810 -1%
CL-FAB5H 787 -0.9%
CL-OP5H 764 -0.9%
CL-OP46H 742 -0.9%
CL-OP47H 733 -1.1%
GEM-38FH 676 -1%
CL-OP47 586 -0.5%
INKL 531 —
CL-OP46 497 —
CL-OP48 490 -0.2%
← Back to feed

Claude Opus 5 Takes Artificial Analysis Intelligence Index #1, Fable 5 Falls to Second

Claude Opus 5 has moved to the top of Artificial Analysis’s Intelligence Index, displacing Fable 5 in the benchmark that tracks a model’s overall capability across reasoning, coding, and knowledge tasks. Artificial Analysis now lists Opus 5 (max) and Opus 5 (xhigh) as the highest-intelligence models available, with Fable 5 (with fallback) second and GPT-5.6 Sol (max) third.

Fable 5 had held the Intelligence Index summit since its June 9 launch at 64.9 — the first time any model had cleared 64 on AA’s scale. The index incorporates AA’s GDPval evaluation and several independent task suites, weighted to reflect real-world professional use. Opus 5’s climb above 64.9 completes a sweep of the major third-party evaluations.

The Benchmark Sweep

Opus 5 now holds the leading position across four independent evaluations:

  • SWE-bench Verified: 97.0% (vals.ai, July 31) — first model to hit the ceiling used by Stack Futures to normalise agentic scores
  • AA Intelligence Index: #1, above Fable 5 at 64.9
  • AA-Briefcase: Elo 1720, 146 points clear of Fable 5
  • LiveBench Agentic Coding: 65.2%, three points ahead of Fable 5 Max Effort (62.2%)

On ARC-AGI-3, Opus 5 posted 30.2% — a 4x improvement over the prior record. On Frontier-Bench, it set SOTA at roughly half the cost of Fable 5 per task.

The Cost Picture

Fable 5 is priced at $15/$75 per million tokens. Opus 5 Fast, which retains full capability at a reduced compute budget, runs at $10/$50 — a third cheaper on input and 33% cheaper on output. On AA-Briefcase, Opus 5 completed tasks 20% cheaper per task than Fable 5 despite outscoring it by 146 Elo points.

The pricing divergence matters for agentic workloads. As token counts compound across multi-step agent runs, even modest per-token differences accumulate quickly. Fable 5 remains competitive on overall LiveBench score (83.0, still above Opus 5’s 80.1), but it costs $1.439 per LiveBench task versus Opus 5’s $0.699.

What This Means for the Leaderboard

The AA Intelligence Index has now had three #1 holders since May: GPT-5.5 through late June, Fable 5 from June 9 through late July, and Opus 5 from this week. The frontier is compressing — Fable 5, Sol, and Opus 5 are separated by a fraction of the full index range — but the direction of movement is clear.

Anthropic holds the top two positions: Opus 5 at #1, Fable 5 at #2. For the first time since GPT-5.4 launched in early 2026, no OpenAI model sits at the summit of AA’s main intelligence benchmark.

GPT-5.6 Sol remains at #3 and still posts the highest Reasoning score on LiveBench (91.7), suggesting its strongest advantage is in pure logical deduction rather than the multi-task knowledge and coding weighting of the AA Intelligence Index.