Which is better, Claude Fable 5 or Opus 5?

Across 47 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 18 and Opus 5 leads on 29 — Opus 5 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude Fable 5 vs Opus 5

AnthropicvsAnthropic47 shared benchmarks18–29 head-to-head

Benchmark	Claude Fable 5	Opus 5
AA Agentic Index	56.6	59.2
AA Intelligence	62.1	63.1
AA-LCR	76.7	78.7
AA-Omniscience	43.3	37.1
arc_agi_1	90.5	97.5
ARC-AGI-2	76.8	90.4
arena_elo	1508	1487
arena_text_factuality	1486	1481
Artificial Analysis Coding Index	76.5	78
browsecomp	88	90.8
coding_arena_elo	1626	1662
critpt	28.6	29.1
DeepSWE 1.1	70	68.8
Finance Agent v2	56.3	58.6
GDP (Surge AI)	29.8	85.5
gdpval	61.9	67.2
GDPval-AA (Elo)	1932	1827
GDPval-AA v2	1760	1861
GPQA Diamond	92.6	93.7
healthbench_professional	66	59.8
HiL-Bench	56.3	57
HLE	64.5	64.7
HLE (with tools)	64.5	64.7
Legal Agent Benchmark	13.3	23
livebench_agentic_coding	62.2	65.2
livebench_coding	86	81.5
livebench_data_analysis	80.5	74.5
livebench_instruction_following	75.8	63.8
livebench_language	90.7	88.7
livebench_math	96	95.7
livebench_reasoning	89.7	91.2
MCP Atlas	84.7	85.8
MMMU-Pro	84.2	84.7
officeqa_pro	69.9	66.9
OmniScience Accuracy	65.3	60.9
OmniScience Non-Hallucination	36.4	40.5
OSWorld 2.0	66.1	70.6
OSWorld-Verified	85	4
scicode	60.2	55.7
simplebench	81.9	80.6
simpleqa_verified	68.3	56.7
swe_bench_multimodal	54.1	59.4
SWE-bench Multilingual	86.6	89.5
SWE-bench Pro	80	79.2
SWE-bench Verified	95	96
TauBench V3 - Banking	38.1	44.7
Terminal-Bench 2.1	88	89.1

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.