Which is better, GPT-5.1 or GPT-5.5?

Across 48 shared benchmarks tracked on Vector Wire, GPT-5.1 leads on 5 and GPT-5.5 leads on 42 — 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-5.1 vs GPT-5.5

OpenAIvsOpenAI48 shared benchmarks5–42 head-to-head

Benchmark	GPT-5.1	GPT-5.5
AA Agentic Index	32.2	47.4
AA Intelligence	37.5	56.3
AA-LCR	76.7	79
AA-Omniscience	5.4	20.5
arc_agi_1	72.8	95
ARC-AGI-2	17.6	85
arena_elo	1439	1482
arena_text_factuality	1445	1482
arena_vision	1250	1283
Artificial Analysis Coding Index	49.4	74.9
browsecomp	50.8	84.4
coding_arena_elo	1421	1457
critpt	4.9	27.1
Fortress	25.7	16.3
frontiermath_tier_4	12.5	35.4
gdpval	25	49.5
GPQA Diamond	88.1	93.6
HealthBench Consensus length-adjusted	95	95.6
HealthBench Hard length-adjusted	25.4	31.5
HealthBench Professional length-adjusted	39.6	51.8
healthbench_length_adjusted	50.9	56.5
HLE	28.5	52.2
HLE (with tools)	42.7	52.2
hmmt_nov_2025	91.7	96.5
IFBench	72.9	75.9
Image input eval - extremism (not_unsafe)	1	1
Image input eval - harms-erotic (not_unsafe)	1	1
Image input eval - hate (not_unsafe)	1	1
Image input eval - self-harm (not_unsafe)	1	1
matharena_visual_math_overall	76.9	94.9
MCP Atlas	50.1	82.8
MMMU-Pro	75.5	83.2
OmniScience Accuracy	37.7	58
OmniScience Non-Hallucination	48.1	12
scicode	43.3	56.1
simplebench	53.2	69
simpleqa_verified	48.9	63.1
SWE-bench Verified	76.3	80.6
TauBench V3 - Banking	15.9	39
Terminal-Bench 2.0	47.6	82.7
Terminal-Bench 2.1	52.4	84.3
Terminal-Bench Hard	45.5	60.6
vectara_answer_rate	100	100
vectara_avg_summary_length	254.4	129.6
vectara_factual_consistency	89.1	90.7
vectara_hallucination_rate ↓	12.1	9.3
τ²-Bench Telecom (AA run)	81.9	93.9
τ³-Bench	14	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.