Which is better, Claude Fable 5 or Claude Opus 4.8?

Across 80 shared benchmarks tracked on Vector Wire, Claude Fable 5 leads on 70 and Claude Opus 4.8 leads on 10 — 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 4.8

AnthropicvsAnthropic80 shared benchmarks70–10 head-to-head

Benchmark	Claude Fable 5	Claude Opus 4.8
AA Agentic Index	56.6	49.4
AA Intelligence	62.1	57.3
AA-LCR	76.7	73
AA-Omniscience	43.3	28.8
Agents' Last Exam	25.7	27
AndroidBench	84.5	69.8
apex_agents	59.2	39.4
arc_agi_1	90.5	92.5
ARC-AGI-2	76.8	72.1
arena_elo	1508	1482
arena_text_factuality	1486	1464
arena_vision	1307	1280
Artificial Analysis Coding Index	76.5	74.3
baby_vision_with_python	90.5	81.2
browsecomp	88	88.5
charxiv_rq	89.4	80.5
coding_arena_elo	1626	1539
CorpFin v2	71.8	66.7
critpt	28.6	20.9
deepsearchqa_f1	94.2	93.1
DeepSWE	70	59
DeepSWE 1.1	70	59
Finance Agent v2	56.3	53.9
GDP (Surge AI)	29.8	84.8
gdpval	61.9	54.2
GDPval-AA (Elo)	1932	1890
GDPval-AA v2	1760	1593
GDPval-AA v2 (Elo)	1747	1593
GPQA Diamond	92.6	93.6
healthbench_professional	66	57.4
HiL-Bench	56.3	35.3
HLE	64.5	57.9
HLE (with tools)	64.5	57.9
IFBench	63.5	62.2
JobBench	57.4	48.4
Legal Agent Benchmark	13.3	10
Legal Research Bench	49.5	43.8
livebench	78.3	77.2
livebench_agentic_coding	62.2	50.5
livebench_coding	86	81.8
livebench_data_analysis	80.5	66
livebench_instruction_following	75.8	72
livebench_language	90.7	79.7
livebench_math	96	94.3
livebench_reasoning	89.7	89.2
mathvision	94.8	86.7
MCP Atlas	84.7	83.6
MCP Mark Verified	87.4	76.4
MLS Bench Lite	49.9	42.8
MMMU-Pro	84.2	78.9
officeqa_pro	69.9	66.2
OmniDocBench	89.8	87.9
OmniScience Accuracy	65.3	48.8
OmniScience Non-Hallucination	36.4	60.7
OSWorld 2.0	66.1	55.7
OSWorld-Verified	85	83.4
PaperBench	88.8	80.3
PerceptionBench	57.2	47.2
PostTrainBench	41.4	37.2
Program Bench	76.8	71.9
scicode	60.2	53.5
simplebench	81.9	64.8
simpleqa_verified	68.3	39.5
SpreadsheetBench 2	34.7	31.6
swe_bench_multimodal	54.1	38.4
SWE-bench Multilingual	86.6	84.4
SWE-bench Pro	80	69.2
SWE-bench Verified	95	88.6
SWE-Marathon	35	40
Tau 3 Banking	26.8	27.6
TauBench V3 - Banking	38.1	34.2
Terminal-Bench 2.1	88	85
Terminal-Bench Hard	62.9	58.3
toolathlon	61.7	59.9
Toolathlon Avg turns	19.8	24.5
Toolathlon Pass@∞	55.6	48.1
WideSearch	81.2	72.9
WorldVQA ForceAnswer	56.7	39.1
τ²-Bench Telecom (AA run)	98.5	94.4
τ³-Bench	26.8	27.6

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.