Which is better, Claude Fable 5 or Claude Sonnet 5?

Across 44 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 40 and Claude Sonnet 5 leads on 4 — Claude Fable 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 Claude Sonnet 5

AnthropicvsAnthropic44 shared benchmarks40–4 head-to-head

Benchmark	Claude Fable 5	Claude Sonnet 5
AA Agentic Index	56.6	49.7
AA Intelligence	62.1	55.3
AA-LCR	76.7	77
AA-Omniscience	43.3	16.4
arena_elo	1508	1463
arena_text_factuality	1486	1453
Artificial Analysis Coding Index	76.5	71.5
browsecomp	88	84.7
charxiv_rq	89.4	77
coding_arena_elo	1626	1540
critpt	28.6	16.9
DeepSWE 1.1	70	54
GDP (Surge AI)	29.8	81.6
gdpval	61.9	54.8
GDPval-AA v2	1760	1618
GPQA Diamond	92.6	91.1
healthbench_professional	66	57.8
HLE	64.5	57.4
Legal Agent Benchmark	13.3	5.8
livebench_agentic_coding	62.2	59.4
livebench_coding	86	80.7
livebench_data_analysis	80.5	71.7
livebench_instruction_following	75.8	63.9
livebench_language	90.7	75
livebench_math	96	92.9
livebench_reasoning	89.7	88.7
MMMU-Pro	84.2	77.3
officeqa_pro	69.9	59.4
OmniScience Accuracy	65.3	40
OmniScience Non-Hallucination	36.4	60.6
OSWorld-Verified	85	81.2
scicode	60.2	53.6
simplebench	81.9	60.6
simpleqa_verified	68.3	25
swe_bench_multimodal	54.1	28.1
SWE-bench Multilingual	86.6	78.3
SWE-bench Pro	80	63.2
SWE-bench Verified	95	85.2
TauBench V3 - Banking	38.1	37.3
Terminal-Bench 2.1	88	80.5
Terminal-bench 3.0	34.1	14.6
toolathlon	61.7	54.3
Toolathlon Pass@∞	55.6	40.7
τ³-Bench	26.8	28.2

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.