Which is better, Claude Fable 5 or GPT-5.2?

Across 46 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 41 and GPT-5.2 leads on 5 — 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 GPT-5.2

AnthropicvsOpenAI46 shared benchmarks41–5 head-to-head

Benchmark	Claude Fable 5	GPT-5.2
AA Agentic Index	56.6	60.2
AA Intelligence	62.1	43.3
AA-LCR	76.7	79.3
AA-Omniscience	43.3	0.3
arc_agi_1	90.5	86.2
ARC-AGI-2	76.8	52.9
arena_elo	1508	1437
arena_text_factuality	1486	1442
arena_vision	1307	1245
Artificial Analysis Coding Index	76.5	48.7
browsecomp	88	65.8
charxiv_rq	89.4	82.1
coding_arena_elo	1626	1418
critpt	28.6	11.6
deepsearchqa_f1	94.2	71.3
gdpval	61.9	48.3
GDPval-AA (Elo)	1932	1462
GPQA Diamond	92.6	92.4
HLE	64.5	45.5
HLE (with tools)	64.5	45.5
IFBench	63.5	75.4
livebench	78.3	74.8
livebench_agentic_coding	62.2	50.3
livebench_coding	86	76.1
livebench_data_analysis	80.5	78.2
livebench_instruction_following	75.8	61.8
livebench_language	90.7	79.8
livebench_math	96	93.2
livebench_reasoning	89.7	83.2
mathvision	94.8	83
MCP Atlas	84.7	68
MMMU-Pro	84.2	79.5
OmniScience Accuracy	65.3	44.3
OmniScience Non-Hallucination	36.4	38.4
PaperBench	88.8	63.7
scicode	60.2	52.1
simplebench	81.9	45.8
simpleqa_verified	68.3	38.9
SWE-bench Multilingual	86.6	72
SWE-bench Pro	80	55.6
SWE-bench Verified	95	80
TauBench V3 - Banking	38.1	11.1
Terminal-Bench Hard	62.9	62.2
toolathlon	61.7	46.3
τ²-Bench Telecom (AA run)	98.5	98.7
τ³-Bench	26.8	12

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.