Which is better, GPT-5 or GPT-5.5?

Across 43 shared benchmarks tracked on Vector Wire, GPT-5 leads on 7 and GPT-5.5 leads on 35 — GPT-5.5 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

GPT-5 vs GPT-5.5

OpenAIvsOpenAI43 shared benchmarks7–35 head-to-head

Benchmark	GPT-5	GPT-5.5
AA Agentic Index	49.7	47.4
AA Intelligence	35.3	56.3
AA-LCR	76.3	79
AA-Omniscience	-8.7	20.5
arc_agi_1	56.2	95
ARC-AGI-2	7.5	85
arena_vision	1212	1283
Artificial Analysis Coding Index	38.9	74.9
browsecomp	54.9	84.4
coding_arena_elo	1419	1457
critpt	5.7	27.1
Fortress	17	16.3
frontiermath_tier_4	12.5	35.4
gdpval	32.5	49.5
GPQA Diamond	87.3	93.6
HealthBench	67.2	56.5
HealthBench Consensus length-adjusted	95.6	95.6
HealthBench Hard length-adjusted	34.7	31.5
HealthBench Professional length-adjusted	46.2	51.8
healthbench_length_adjusted	57.7	56.5
HLE	28.5	52.2
HLE (with tools)	35.2	52.2
hmmt_nov_2025	89.2	96.5
IFBench	73.1	75.9
matharena_visual_math_overall	78.8	94.9
MMMU-Pro	78.4	83.2
OmniScience Accuracy	40.3	58
OmniScience Non-Hallucination	21	12
scicode	43	56.1
simplebench	56.7	69
simpleqa_verified	50.6	63.1
SWE-bench Pro	41.8	58.6
SWE-bench Verified	74.9	80.6
TauBench V3 - Banking	22.1	39
Terminal-Bench 2.0	35.2	82.7
Terminal-Bench 2.1	35.2	84.3
Terminal-Bench Hard	37.9	60.6
vectara_answer_rate	99.9	100
vectara_avg_summary_length	162.7	129.6
vectara_factual_consistency	85.3	90.7
vectara_hallucination_rate ↓	14.7	9.3
τ²-Bench Telecom (AA run)	86.5	93.9
τ³-Bench	19.6	31.3

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.