Which is better, Claude Fable 5 or GPT-5.6 Sol?

Across 79 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 48 and GPT-5.6 Sol leads on 31 — 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-5.6 Sol

AnthropicvsOpenAI79 shared benchmarks48–31 head-to-head

Benchmark	Claude Fable 5	GPT-5.6 Sol
AA Agentic Index	56.6	57.8
AA Intelligence	62.1	61
AA-LCR	76.7	77.7
AA-Omniscience	43.3	22
Agents' Last Exam	25.7	52.7
AndroidBench	84.5	74
apex_agents	59.2	56.7
arc_agi_1	90.5	97.5
ARC-AGI-2	76.8	92.5
arena_elo	1508	1482
arena_text_factuality	1486	1474
Artificial Analysis Coding Index	76.5	78.3
baby_vision_with_python	90.5	88.9
browsecomp	88	90.4
BrowseComp (w/ Ctx)	88	89.4
charxiv_rq	89.4	87.8
coding_arena_elo	1626	1619
CorpFin v2	71.8	64.4
critpt	28.6	32.3
CursorBench v3.2	70.5	67.2
DeepSWE	70	73
DeepSWE 1.1	70	73
Finance Agent v2	56.3	53.8
FORTRESS (Adversarial)	96	82.4
FORTRESS (Benign)	55.1	98.1
FrontierCode v1.1 (Extended)	63.6	60.6
GDP (Surge AI)	29.8	30.7
gdpval	61.9	61.1
GDPval-AA v2	1760	1748
GDPval-AA v2 (Elo)	1747	1736
Global-MMLU-Lite	93.3	91.8
GPQA Diamond	92.6	94.6
Harvey LAB (Vals)	11.3	2.5
healthbench_professional	66	60.5
HiL-Bench	56.3	32.3
HLE	64.5	49.5
HLE (with tools)	64.5	58
IFBench	63.5	72.7
JobBench	57.4	45.4
Legal Research Bench	49.5	48.1
livebench_agentic_coding	62.2	56.2
livebench_coding	86	83.9
livebench_data_analysis	80.5	79.8
livebench_instruction_following	75.8	71.8
livebench_language	90.7	87.7
livebench_math	96	96.2
livebench_reasoning	89.7	91.7
mathvision	94.8	95.8
MCP Atlas	84.7	83.6
MCP Mark Verified	87.4	92.9
MLS Bench Lite	49.9	46.2
MMMU-Pro	84.2	84.6
officeqa_pro	69.9	63.2
OmniDocBench	89.8	85.8
OmniScience Accuracy	65.3	59.4
OmniScience Non-Hallucination	36.4	10.6
OSWorld 2.0	66.1	62.6
OSWorld-Verified	85	83
PaperBench	88.8	90.5
PerceptionBench	57.2	59.7
PostTrainBench	41.4	34.6
Program Bench	76.8	77.6
scicode	60.2	56.9
simplebench	81.9	64.8
simpleqa_verified	68.3	71.6
SpreadsheetBench 2	34.7	32.4
strongreject	98.7	98.5
SWE-bench Pro	80	64.6
SWE-Marathon	35	39
SWEBench Pro (Public)	80	64.6
Tau 3 Banking	26.8	33
TauBench V3 - Banking	38.1	44.3
Terminal Bench 2.1 (Best Harness)	84.6	89.5
Terminal-Bench 2.1	88	89.5
Terminal-bench 3.0	34.1	34.6
Terminal-Bench Hard	62.9	65.9
toolathlon	61.7	58
WorldVQA ForceAnswer	56.7	41.8
τ²-Bench Telecom (AA run)	98.5	85.1

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.