Which is better, Claude Opus 4.7 or Claude Opus 4.8?

Across 68 shared benchmarks tracked on Vector Wire, Claude Opus 4.7 leads on 25 and Claude Opus 4.8 leads on 43 — Claude Opus 4.8 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude Opus 4.7 vs Claude Opus 4.8

AnthropicvsAnthropic68 shared benchmarks25–43 head-to-head

Benchmark	Claude Opus 4.7	Claude Opus 4.8
AA Agentic Index	64.6	49.4
AA Intelligence	55	57.3
AA-LCR	75.3	73
AA-Omniscience	27.3	28.8
aime_2026	95.8	100
arc_agi_1	92	92.5
arc_agi_3	0.2	1.5
ARC-AGI-2	75.8	72.1
arena_elo	1494	1482
arena_text_factuality	1474	1464
arena_vision	1304	1280
Artificial Analysis Coding Index	73.6	74.3
browsecomp	79.8	88.5
ChartQAPro (No tools)	67.6	69.4
ChartQAPro (With tools)	69.8	72.3
charxiv_reasoning_with_tools	91	89.9
charxiv_rq	82.1	80.5
coding_arena_elo	1557	1539
critpt	12	20.9
cybergym	73.8	83.1
Finance Agent v2	51.5	53.9
finance_agent	71.5	53.9
frontiermath_tier_4	22.9	31.3
gdpval	58.6	54.2
GDPval-AA (Elo)	1753	1890
GPQA Diamond	94.2	93.6
GraphWalks — Parents task, 256K-token context subset	93.6	99.3
GraphWalks BFS 256K	76.9	85.9
HiL-Bench	41.7	35.3
HLE	54.7	57.9
HLE (with tools)	54.7	57.9
hmmt_feb_2026	93.9	96.7
IFBench	58.6	62.2
Legal Agent Benchmark	7.1	10
livebench	76.9	77.2
livebench_agentic_coding	50.7	50.5
livebench_coding	82.1	81.8
livebench_data_analysis	78.3	66
livebench_instruction_following	66.7	72
livebench_language	77.9	79.7
livebench_math	92.8	94.3
livebench_reasoning	87.2	89.2
MCP Atlas	79.1	83.6
MMMU-Pro	78.8	78.9
officeqa	86.3	77.6
officeqa_pro	80.6	66.2
OmniScience Accuracy	48.9	48.8
OmniScience Non-Hallucination	57.7	60.7
OSWorld-Verified	82.8	83.4
scicode	54.5	53.5
screenspot_pro_no_tools	79.5	87.9
screenspot_pro_with_tools	87.6	87.9
simplebench	61.7	64.8
simpleqa_verified	50.6	39.5
swe_bench_multimodal	34.5	38.4
SWE-bench Multilingual	80.5	84.4
SWE-bench Pro	64.3	69.2
SWE-bench Verified	87.6	88.6
TauBench V3 - Banking	34.6	34.2
Terminal-Bench 2.0	69.4	74.6
Terminal-Bench 2.1	83.1	85
Terminal-Bench Hard	54.5	58.3
toolathlon	59.3	59.9
Toolathlon Avg turns	25.9	24.5
Toolathlon Pass@∞	52.8	48.1
usamo_2026	69.3	96.7
τ²-Bench Telecom (AA run)	88.6	94.4
τ³-Bench	28.9	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.