Which is better, Claude Fable 5 or Claude Opus 4.7?

Across 53 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 46 and Claude Opus 4.7 leads on 7 — 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 Opus 4.7

AnthropicvsAnthropic53 shared benchmarks46–7 head-to-head

Benchmark	Claude Fable 5	Claude Opus 4.7
AA Agentic Index	56.6	64.6
AA Intelligence	62.1	55
AA-LCR	76.7	75.3
AA-Omniscience	43.3	27.3
arc_agi_1	90.5	92
ARC-AGI-2	76.8	75.8
arena_elo	1508	1494
arena_text_factuality	1486	1474
arena_vision	1307	1304
Artificial Analysis Coding Index	76.5	73.6
Blueprint-Bench 2	38.6	24.5
browsecomp	88	79.8
charxiv_rq	89.4	82.1
coding_arena_elo	1626	1557
critpt	28.6	12
Finance Agent v2	56.3	51.5
gdpval	61.9	58.6
GDPval-AA (Elo)	1932	1753
GPQA Diamond	92.6	94.2
HiL-Bench	56.3	41.7
HLE	64.5	54.7
HLE (with tools)	64.5	54.7
IFBench	63.5	58.6
Legal Agent Benchmark	13.3	7.1
livebench	78.3	76.9
livebench_agentic_coding	62.2	50.7
livebench_coding	86	82.1
livebench_data_analysis	80.5	78.3
livebench_instruction_following	75.8	66.7
livebench_language	90.7	77.9
livebench_math	96	92.8
livebench_reasoning	89.7	87.2
MCP Atlas	84.7	79.1
MMMU-Pro	84.2	78.8
officeqa_pro	69.9	80.6
OmniScience Accuracy	65.3	48.9
OmniScience Non-Hallucination	36.4	57.7
OSWorld-Verified	85	82.8
scicode	60.2	54.5
simplebench	81.9	61.7
simpleqa_verified	68.3	50.6
swe_bench_multimodal	54.1	34.5
SWE-bench Multilingual	86.6	80.5
SWE-bench Pro	80	64.3
SWE-bench Verified	95	87.6
TauBench V3 - Banking	38.1	34.6
Terminal-Bench 2.1	88	83.1
Terminal-Bench Hard	62.9	54.5
toolathlon	61.7	59.3
Toolathlon Avg turns	19.8	25.9
Toolathlon Pass@∞	55.6	52.8
τ²-Bench Telecom (AA run)	98.5	88.6
τ³-Bench	26.8	28.9

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.