Which is better, Claude Fable 5 or GLM-5.2 Full Open Source?

Across 57 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 50 and GLM-5.2 Full Open Source 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 GLM-5.2 Full Open Source

AnthropicvsZ.ai57 shared benchmarks50–4 head-to-head

Benchmark	Claude Fable 5	GLM-5.2 Full Open Source
AA Agentic Index	56.6	45.7
AA Intelligence	62.1	52.6
AA-LCR	76.7	76.7
AA-Omniscience	43.3	4.4
Agents' Last Exam	25.7	23.8
apex_agents	59.2	35.6
arc_agi_1	90.5	77
ARC-AGI-2	76.8	22.8
arena_elo	1508	1471
arena_text_factuality	1486	1460
Artificial Analysis Coding Index	76.5	68.8
coding_arena_elo	1626	1582
critpt	28.6	20.9
DeepSWE	70	46.2
DeepSWE 1.1	70	44
FORTRESS (Adversarial)	96	71.3
FORTRESS (Benign)	55.1	90
gdpval	61.9	50.3
GDPval-AA v2	1760	1514
GDPval-AA v2 (Elo)	1747	1510
Global-MMLU-Lite	93.3	89.2
GPQA Diamond	92.6	91.2
HiL-Bench	56.3	43.7
HLE	64.5	54.7
HLE (with tools)	64.5	54.7
IFBench	63.5	73.3
JobBench	57.4	43.4
livebench_agentic_coding	62.2	51.8
livebench_coding	86	79.7
livebench_data_analysis	80.5	73.7
livebench_instruction_following	75.8	62.3
livebench_language	90.7	76.2
livebench_math	96	89.8
livebench_reasoning	89.7	78.6
MCP Atlas	84.7	82.6
MLS Bench Lite	49.9	40.4
officeqa_pro	69.9	41.4
OmniScience Accuracy	65.3	24.3
OmniScience Non-Hallucination	36.4	73.7
PostTrainBench	41.4	34.3
Program Bench	76.8	63.7
scicode	60.2	50.5
simplebench	81.9	58.8
simpleqa_verified	68.3	38.1
SpreadsheetBench 2	34.7	28.1
strongreject	98.7	98.5
SWE-bench Pro	80	62.1
SWE-Marathon	35	13
SWEBench Pro (Public)	80	62.1
Tau 3 Banking	26.8	26.8
TauBench V3 - Banking	38.1	34.6
Terminal Bench 2.1 (Best Harness)	84.6	82.7
Terminal-Bench 2.1	88	82.7
Terminal-Bench Hard	62.9	50.8
toolathlon	61.7	48.2
τ²-Bench Telecom (AA run)	98.5	99.1
τ³-Bench	26.8	26.8

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.