Which is better, Claude Opus 4.8 or Gemini 3 Pro?

Across 50 shared benchmarks tracked on Vector Wire, Claude Opus 4.8 leads on 34 and Gemini 3 Pro leads on 15 — 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 Pro

AnthropicvsGoogle50 shared benchmarks34–15 head-to-head

Benchmark	Claude Opus 4.8	Gemini 3 Pro
AA Agentic Index	49.4	52
AA Intelligence	57.3	40.6
AA-LCR	73	73
AA-Omniscience	28.8	15.3
aime_2026	100	91.7
arc_agi_1	92.5	75
ARC-AGI-2	72.1	54
arena_elo	1482	1485
arena_text_factuality	1464	1481
arena_vision	1280	1290
Artificial Analysis Coding Index	74.3	46.5
browsecomp	88.5	59.2
charxiv_rq	80.5	81.4
coding_arena_elo	1539	1438
critpt	20.9	9.1
cybergym	83.1	39.9
deepsearchqa_f1	93.1	63.2
Fortress	18.2	41.7
frontiermath_tier_4	31.3	18.8
gdpval	54.2	34.2
GDPval-AA (Elo)	1890	1195
GPQA Diamond	93.6	91.9
HLE	57.9	45.8
HLE (with tools)	57.9	45.8
hmmt_feb_2026	96.7	86.4
hmmt_nov_2025	96.5	93.3
IFBench	62.2	70.4
imo_answer_bench	83.5	83.3
livebench	77.2	73.4
longbench_v2	69.1	68.2
matharena_visual_math_overall	81.6	84.2
mathvision	86.7	86.1
MCP Atlas	83.6	70.3
MMMU-Pro	78.9	81
MMVU	79.2	78.9
OmniScience Accuracy	48.8	55.8
OmniScience Non-Hallucination	60.7	10
scicode	53.5	56.1
screenspot_pro_no_tools	87.9	72.7
simplebench	64.8	76.4
simpleqa_verified	39.5	72.9
SWE-bench Multilingual	84.4	68.7
SWE-bench Pro	69.2	43.3
SWE-bench Verified	88.6	78
Terminal-Bench 2.0	74.6	54.2
Terminal-Bench Hard	58.3	56.9
tool_decathlon	59.9	36.4
toolathlon	59.9	36.4
Video-MME	86	88.4
τ²-Bench Telecom (AA run)	94.4	98

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.