Which is better, Claude Opus 4.8 or Gemini 3.5 Flash?

Across 53 shared benchmarks tracked on Vector Wire, Claude Opus 4.8 leads on 33 and Gemini 3.5 Flash leads on 17 — 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 4.8 vs Gemini 3.5 Flash

AnthropicvsGoogle53 shared benchmarks33–17 head-to-head

Benchmark	Claude Opus 4.8	Gemini 3.5 Flash
AA Agentic Index	49.4	70.4
AA Intelligence	57.3	52
AA-LCR	73	81
AA-Omniscience	28.8	21.2
aime_2026	100	95
arc_agi_1	92.5	92.5
ARC-AGI-2	72.1	72.1
arena_elo	1482	1477
arena_text_factuality	1464	1466
Artificial Analysis Coding Index	74.3	70.1
charxiv_reasoning_with_tools	89.9	84.9
charxiv_rq	80.5	84.2
coding_arena_elo	1539	1490
critpt	20.9	13.1
DeepSWE	59	37
DeepSWE 1.1	59	37
Finance Agent v2	53.9	57.9
finance_agent	53.9	57.9
frontiermath_tier_4	31.3	14.6
gdpval	54.2	57.8
GDPval-AA (Elo)	1890	1656
GDPval-AA v2	1593	1357
GPQA Diamond	93.6	92.2
HiL-Bench	35.3	27.7
HLE	57.9	42.7
hmmt_feb_2026	96.7	95.5
IFBench	62.2	76.3
Legal Agent Benchmark	10	0.8
livebench	77.2	75
livebench_agentic_coding	50.5	49
livebench_coding	81.8	78.2
livebench_data_analysis	66	64.9
livebench_instruction_following	72	75.6
livebench_language	79.7	84.6
livebench_math	94.3	88.2
livebench_reasoning	89.2	82
matharena_visual_math_overall	81.6	89.9
MCP Atlas	83.6	83.6
MMMU-Pro	78.9	84.3
OmniScience Accuracy	48.8	51.4
OmniScience Non-Hallucination	60.7	38.2
OSWorld-Verified	83.4	78.4
scicode	53.5	53.1
simplebench	64.8	76.7
simpleqa_verified	39.5	68.4
SWE-bench Pro	69.2	55.1
TauBench V3 - Banking	34.2	32.2
Terminal-Bench 2.0	74.6	76.2
Terminal-Bench 2.1	85	78.7
Terminal-Bench Hard	58.3	46.2
toolathlon	59.9	56.5
τ²-Bench Telecom (AA run)	94.4	95.6
τ³-Bench	27.6	25.4

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.