OpenAIReasoning model

GPT-5.4

Reasoning effort

GPT-5.4 is an OpenAI proprietary reasoning model, released 5 Mar 2026. Its best configuration (xhigh reasoning effort) ranks #54 of 756 on the BenchLeader Index at 64.2 ±2.6, in the upper half. The ± is the point: 111 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (72) and lowest in knowledge (60). At $5.63 per million tokens blended it is among the most expensive ranked models. Output speed of 89 tokens per second puts it faster than most, with a first answer in 192.9 s. It has been measured at 7 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 29 Sept 2026.

Blended price: $5.63/M; $2.50 in · $15.00 out
Output speed: 89 tok/s; measured by Artificial Analysis
First answer: 193 s; first token 1.41 s
Context: 1.1M
Full answer: 199 s; median, reasoning included
Cost per run: –
Released: 5 Mar 2026

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index64
Reasoning70
Coding65
Agents & tools64
Maths66
Knowledge60
Instruction following72
Multimodal62
Long context66
Composite66

Versions

OpenAI has shipped 12 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.

Model	Released	Index	Rank
GPT-5.6	9 Jul 2026	–	–
GPT-5.5xhigh	23 Apr 2026	65.9	#37
GPT-5.4xhighthis page	5 Mar 2026	64.2	#54
GPT-5.2xhigh	11 Dec 2025	60.9	#97
GPT-5.1high	12 Nov 2025	58.0	#155
GPT-5high	7 Aug 2025	58.1	#151
GPT-4.1	14 Apr 2025	48.6	#404
GPT-4.5	27 Feb 2025	49.9	#362
GPT 4 0125	25 Jan 2024	45.7	#485
GPT-4	6 Nov 2023	44.3	#518
GPT 4 0613	13 Jun 2023	39.9	#645
GPT 4 0314	14 Mar 2023	43.5	#543

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
no reasoning	52.7	#294	64 tok/s	0.78 s	$0.0055	Agents & tools 67 · Coding 55 · Composite 50 · Instruction following 50 · Knowledge 55 · Long context 54 · Maths 49 · Multimodal 54 · Reasoning 49
low	59.9	#114	60 tok/s	2.30 s	$0.0055	Agents & tools 71 · Composite 61 · Instruction following 65 · Knowledge 64 · Long context 64 · Maths 61 · Multimodal 61 · Reasoning 54
medium	55.2	#218	50 tok/s	1.41 s	$0.0055	Coding 49 · Maths 66 · Reasoning 61
high	58.0	#156	50 tok/s	1.41 s	$0.0055	Agents & tools 49 · Coding 55 · Human preference 67 · Knowledge 52 · Maths 68 · Multimodal 65 · Reasoning 61
xhighbest	64.2	#54	89 tok/s	193 s	$0.0055	Agents & tools 64 · Coding 65 · Composite 66 · Instruction following 72 · Knowledge 60 · Long context 66 · Maths 66 · Multimodal 62 · Reasoning 70
thinking	–	–	50 tok/s	1.41 s	$0.0055	Knowledge 70 · Multimodal 65
not stated	58.6	#143	50 tok/s	1.41 s	$0.0055	Agents & tools 60 · Coding 51 · Human preference 64 · Maths 65 · Multimodal 64 · Reasoning 62

How its index has moved

8 Sept 2026 to 5 Oct 2026

xhigh
low
not stated
high

The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
GPQA Diamond	74.8%#146	84.8%#87	88.9%#51	89.9%#45	93.3%#18	–	–	Epoch AI Benchmarking Hub
Humanity's Last Exam	–	–	–	–	36.2%#8	–	–	Scale AI / CAIS
LMArena Hard Prompts	–	–	–	1497#37	–	–	1487#55	LMArena
LiveBench Reasoningnot in index	–	–	–	–	88.1%#23	–	–	LiveBench
GPQA Diamond (AA)not in index	74.8%#248	87.1%#99	–	–	92.0%#41	–	–	Artificial Analysis
Humanity's Last Exam (AA)not in index	11.3%#275	30.8%#127	–	–	43.7%#50	–	–	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	–	–	91.7%#18	–	–	Vals AI
Kagi LLM Benchmark	–	–	–	–	–	–	63.8%#40	Kagi LLM Benchmark
ARC-AGI-1	–	68.2%#124	86.2%#87	92.7%#46	93.7%#42	–	–	ARC Prize
ARC-AGI-2	–	29.2%#115	55.4%#85	67.5%#58	74.0%#47	–	–	ARC Prize
ARC-AGI-3	–	–	–	0.2%#42	–	–	–	ARC Prize
CritPt	0.6%#234	7.4%#125	–	–	23.4%#46	–	–	Artificial Analysis
Chess Puzzlesnot in index	5.0%#156	20.0%#85	38.0%#32	38.0%#32	44.0%#18	–	–	Epoch AI Benchmarking Hub
Mystery Game Puzzlesnot in index	16.0%#78	17.0%#70	28.0%#34	–	37.0%#15	–	–	Epoch AI Benchmarking Hub
CL-benchnot in index	–	–	–	–	27.9%#1	–	–	Epoch AI Benchmarking Hub
CL-bench Lifenot in index	–	–	–	19.3%#3	21.7%#2	–	13.8%#7	Epoch AI Benchmarking Hub
LMCAnot in index	–	–	–	–	52.0%#43	–	–	Epoch AI Benchmarking Hub
DTBenchnot in index	–	–	–	–	94.4%#44	–	–	Epoch AI Benchmarking Hub
ForecastBenchnot in index	–	–	–	–	–	–	59.5%#40	ForecastBench

Coding

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
SWE-bench Verified (Epoch)	–	–	–	76.9%#8	–	–	–	Epoch AI Benchmarking Hub
SciCode	–	–	–	–	56.6%#30	–	–	SciCode
WeirdML	57.4%#53	–	–	–	77.7%#24	–	–	WeirdML
GSO-Bench	–	–	–	25.5%#15	31.4%#12	–	–	GSO-Bench
LMArena Coding	–	–	–	1521#34	–	–	1511#57	LMArena
LMArena WebDev	–	–	1444#66	1465#61	–	–	1395#83	LMArena
LiveBench Codingnot in index	–	–	–	–	77.5%#35	–	–	LiveBench
LiveCodeBench	–	–	–	–	84.1%#39	–	–	Vals AI
SWE-bench (Vals)not in index	–	–	–	–	78.2%#30	–	–	Vals AI
SWE-Bench Pro	–	–	–	–	59.1%#2	–	–	Scale AI SEAL
Code Migrationnot in index	–	–	–	–	35.0%#31	–	–	Vals AI
ProgramBenchnot in index	–	–	–	0.5%#13	0.0%#23	–	–	Vals AI
Vibe Code Bench v1.1not in index	–	–	–	–	67.4%#35	–	48.5%#50	Vals AI
SWE-Bench Pro (private)not in index	–	–	–	–	59.1%#2	–	–	Scale AI SEAL
SWE Atlas: Codebase QnAnot in index	–	–	–	–	40.8%#10	–	–	Scale AI SEAL
SWE Atlas: Refactoringnot in index	–	–	–	–	44.3%#8	–	–	Scale AI SEAL
SWE Atlas: Test Writingnot in index	–	–	–	–	44.4%#8	–	–	Scale AI SEAL
MirrorCodenot in index	–	–	–	15.6%#7	–	–	–	Epoch AI Benchmarking Hub
DeepSWEnot in index	–	–	–	–	51.8%#46	–	–	Epoch AI Benchmarking Hub
ALE-Benchnot in index	1086.0#46	–	1520.7#18	1607#13	–	–	–	Epoch AI Benchmarking Hub
AlgoTunenot in index	–	–	–	1.9#3	–	–	–	AlgoTune
Terminal-Bench 2.1 (AA)not in index	–	–	–	–	78.3%#52	–	–	Artificial Analysis

Agents & tools

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
Terminal-Bench	–	–	–	–	–	–	81.8%#2	Terminal-Bench
APEX-Agents	–	–	–	–	–	–	52.4%#25	Mercor
LMArena Agent	–	–	–	1.1#27	–	–	–	LMArena
LiveBench Agentic Codingnot in index	–	–	–	–	53.8%#33	–	–	LiveBench
Terminal-Bench Hard	37.9%#63	43.2%#41	–	–	57.6%#11	–	–	Artificial Analysis
τ²-Bench Telecom (AA)not in index	36.0%#233	74.6%#139	–	–	87.1%#80	–	–	Artificial Analysis
MCP Atlas	–	–	–	–	70.6%#19	–	–	Scale AI SEAL
HiL-Bench	–	–	–	–	–	–	9.7%#15	Scale AI SEAL
GDPval (AA)not in index	–	–	–	–	36.6%#100	–	–	Artificial Analysis
τ²-Bench Banking (AA)	–	–	–	–	39.6%#31	–	–	Artificial Analysis
APEX-Agents (AA)not in index	–	–	–	–	33.3%#9	–	–	Artificial Analysis
Harvey's Legal Agent Benchmarknot in index	–	–	–	–	0.0%#53	–	–	Vals AI
SkillsBenchnot in index	–	–	–	–	51.7%#19	–	–	Vals AI
Terminal-Bench 2.0 (Vals)not in index	–	–	–	58.4%#10	–	–	–	Vals AI
DeepResearch Benchnot in index	–	35.1%#39	–	–	–	–	–	DeepResearch Bench
PostTrainBenchnot in index	–	–	–	19.0%#12	–	–	–	Epoch AI Benchmarking Hub
METR Time Horizonsnot in index	–	–	–	–	74.3%#7	–	74.3%#7	METR
GBAEvalnot in index	–	–	–	–	–	–	45.1%#10	Epoch AI Benchmarking Hub
Vending-Bench 2not in index	–	–	–	–	–	–	6144.2#19	Andon Labs
DeepSearchQAnot in index	–	–	–	–	–	–	63.7%#3	Google

Maths

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
FrontierMath Tiers 1–3	–	–	–	–	78.6%#17	–	–	Epoch AI Benchmarking Hub
FrontierMath Tier 4	–	–	–	–	49.0%#24	–	–	Epoch AI Benchmarking Hub
OTIS Mock AIME	57.8%#159	84.4%#90	95.6%#36	97.8%#24	95.3%#44	–	–	Epoch AI Benchmarking Hub
ProofBench	–	–	–	–	56.0%#22	–	–	Vals AI
LiveBench Mathematicsnot in index	–	–	–	–	94.2%#19	–	–	LiveBench
AIME (Vals)	–	–	–	–	96.7%#5	–	–	Vals AI
AIME 2026	–	–	–	–	99.2%#3	–	–	MathArena
HMMT February 2026	–	–	–	–	97.7%#2	–	–	MathArena
MathArena Apex	–	–	–	–	54.2%#6	–	–	MathArena
LMArena Maths	–	–	–	1492#25	–	–	1460#60	LMArena

Knowledge

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
SimpleQA Verified	–	–	–	–	45.1%#37	–	–	Epoch AI Benchmarking Hub
LiveBench Data Analysisnot in index	–	–	–	–	79.3%#14	–	–	LiveBench
AA-Omniscience	-15.7#207	4.8#100	–	–	5.8#96	–	–	Artificial Analysis
MMLU-Pro	–	–	–	–	87.5%#27	–	–	Vals AI
LegalBench	–	–	–	–	86.0%#14	–	–	Vals AI
CorpFin	–	–	–	–	65.3%#32	–	–	Vals AI
TaxEval	–	–	–	–	74.0%#42	–	–	Vals AI
MedQA	–	–	–	–	96.1%#5	–	–	Vals AI
PRBench Finance	–	–	–	45.6%#18	–	–	–	Scale AI SEAL
PRBench Legal	–	–	–	44.4%#18	–	–	–	Scale AI SEAL
MultiNRC	–	–	–	–	–	58.3%#7	–	Scale AI SEAL
CaseLaw v2not in index	–	–	–	–	63.8%#16	–	–	Vals AI
MedCodenot in index	–	–	–	–	41.3%#52	–	–	Vals AI
MedScribenot in index	–	–	–	–	77.5%#64	–	–	Vals AI
EBR-benchnot in index	–	–	–	–	25.4%#12	–	–	Epoch AI Benchmarking Hub
AA-Omniscience: accuracynot in index	37.4%#136	47.9%#71	–	–	50.9%#59	–	–	Artificial Analysis
AA-Omniscience: non-hallucinationnot in index	15.3%#318	17.4%#299	–	–	8.3%#441	–	–	Artificial Analysis

Instruction following

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
LiveBench Languagenot in index	–	–	–	–	82.6%#24	–	–	LiveBench
LiveBench Instruction Followingnot in index	–	–	–	–	70.2%#28	–	–	LiveBench
IFBench	48.4%#181	65.9%#100	–	–	74.0%#37	–	–	Artificial Analysis
LMArena Instruction Followingnot in index	–	–	–	1471#36	–	–	1460#51	LMArena

Human preference

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
LMArena Text	–	–	–	1475#34	–	–	1465#53	LMArena
EQ-Bench 4	–	–	–	–	–	–	1272#7	EQ-Bench
LMArena Creative Writingnot in index	–	–	–	1443#55	–	–	1435#65	LMArena
LMArena Multi-turnnot in index	–	–	–	1493#18	–	–	1480#40	LMArena
LMArena Longer Queriesnot in index	–	–	–	1482#42	–	–	1477#50	LMArena

Multimodal

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
LMArena Vision	–	–	–	1302#17	–	–	1292#25	LMArena
MMMU-Pro	70.6%#135	78.0%#74	–	–	78.4%#71	–	–	Artificial Analysis
VISTA	–	–	–	–	–	50.9%#8	–	Scale AI SEAL
MMMU-Pro (Vals)not in index	–	–	–	–	87.5%#14	–	–	Vals AI
MortgageTaxnot in index	–	–	–	–	68.3%#17	–	–	Vals AI
SAGEnot in index	–	–	–	–	43.3%#49	–	–	Vals AI
VTBnot in index	–	–	–	29.2%#2	–	–	–	Scale AI SEAL
Blueprint-Bench 2not in index	–	–	–	–	–	–	27.1%#19	Epoch AI Benchmarking Hub

Long context

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source	Trend
AA-LCR	58.3%#269	76.7%#126	–	–	82.0%#41	–	–	Artificial Analysis
LMArena Documentnot in index	–	–	–	–	–	–	1470#18	LMArena

Composite

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source
Epoch Capabilities Indexnot in index	–	–	–	–	–	–	156.9#16	Epoch AI Benchmarking Hub
LiveBench	–	–	–	–	78.0%#18	–	–	LiveBench
AA Intelligence Index	18.2#237	27.6#126	–	–	39.0#65	–	–	Artificial Analysis

Safety & honesty

Benchmark	no reasoning	low	medium	high	xhigh	thinking	not stated	Source	Trend
MASKnot in index	–	–	–	–	–	89.7%#8	–	Scale AI SEAL

Where it wins

Benchmarks where this configuration ranks in the top five of every configuration measured.

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Amazon Bedrock	147 tok/s	0.70 s	$2.75	$16.50	1.1M	–
Azure	113 tok/s	0.62 s	$2.75	$16.50	1.1M	–
OpenAI	91 tok/s	1.47 s	$5.00	$30.00	1.1M	–
Azure	68 tok/s	1.95 s	$2.50	$15.00	1.1M	–
OpenAI	50 tok/s	1.04 s	$2.50	$15.00	1.1M	–
OpenAI	33 tok/s	1.74 s	$1.25	$7.50	1.1M	–

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

input outputnow $2.75 in · $16.50 out

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.250 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0055	$0.0048	3.3 min
Summarise a 30-page report	12,000 / 600	$0.039	$0.019	3.3 min
Code edit	6,000 / 1,500	$0.037	$0.027	3.5 min
Agentic coding session	60,000 / 4,000	$0.210	$0.109	4.0 min
Structured extraction	2,000 / 200	$0.0080	$0.0046	3.3 min