Models / GPT-5.4

OpenAI

GPT-5.4

GPT-5.4 is best for writing, and reasoning and knowledge; pick it when quality matters more than cost. The cheapest model of similar intelligence is MiMo-V2.6-Flash, at $0.18 per million tokens against $5.62.

Superseded. GPT-5.5 is OpenAI's newer model in this line, at $11.25 blended: see GPT-5.5 or compare the two.

Results 16 Nov 2026 · 90 results · 10 sources

Test GPT-5.4 on your prompt → It runs beside the models you could use instead, with every answer checked and costed.

Price per million tokens
$2.50 in · $15.00 out
OpenAI's list price, 10 Nov 2026 · compare prices$0.25 cached input
Speed
Not measured
Intelligence Index
39.0
Artificial Analysis, at xhigh reasoning; 18.2–39.0 across 3 settings
Spring Prompt overall
Not ranked yet
needs our benchmarks and two groups of results
Developer
OpenAI, based in the US
Weights not published
Released
5 Mar 2026
1,050,000 tokens of context

Where it stands among the models you could choose

Available models grouped by price, intelligence, speed and developer location. Retired models are excluded.

AmongIntelligence IndexSpring Prompt overallPrice, blendedSpeed
Models available today 31st of 185Claude Opus 5.5 – $5.62 blended, 90th percentilecheapest: Mistral Nemo –
Models priced $3–10 per million tokens 16th of 38Claude Opus 5.5 – sets this group –
Models with similar intelligence sets this group – $5.62 blended, 84th percentilecheapest: MiMo-V2.6-Flash –
Models from developers based in the US 22nd of 92Claude Opus 5.5 – $5.62 blended, 80th percentilecheapest: Gemma 3 4B –

The name under each rank is the leader of that group; green is the top quarter of the group and red the bottom quarter, by rank. Price is shown as a percentile: the share of the group that costs less, so lower is cheaper. "Sets this group" marks the measure the group is defined by. Spring Prompt overall is our score out of 100 across every source (how it works). Price is the developer's list price (or the typical OpenRouter host where we haven't read one) for three input tokens to one output; intelligence and speed are from Artificial Analysis (data sourced from Artificial Analysis); "similar intelligence" means within 4 points on its Intelligence Index. Developer locations are where each company is based, not where a model is served. Groups under 3 models are not ranked.

How good is it, and for what?

Each use case ranks the available models its benchmarks measured, then those priced $3–10 per million tokens. The bar shows its percentile in that field, best to the right. Open a row for the results behind it.

Product listingsTurning a sparse product feed and photos into listings that can go live Not measured
  • CatalogBench: has not measured this model.
Decks from an analysisTurning a finished analysis into a deck you could present as it is Not measured
  • DeckBench: has not measured this model.
User surveysPlanning a user survey and reading its results without being misled Not measured
  • SurveyBench: has not measured this model.
Marketing planningPlanning a year of ad spend without overspending Not measured
  • ROASBench: has not measured this model.
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 26th of 159 available · 13th of 35 at its price
BenchmarkGPT-5.4Rank among availableBest available
OpenHands IndexOpenHands Index: average · reported 64.3% 6th of 31 Claude Fable 5 81.0%
Microsoft STATE-BenchSTATE-Bench v0.7: task success (pass@1) · reported 55.7% 1st of 4 This model
Artificial AnalysisArtificial Analysis: Terminal-Bench 2.1 · reported 78.3% 27th of 107 Claude Fable 5.1 91.4%
Artificial AnalysisArtificial Analysis: τ-bench banking · reported 39.6% 17th of 104 Grok 4.6 50.7%
Vending-Bench 2Vending-Bench 2: money after a year · reported $6,144 16th of 53 GPT-6 Astra $15,515
  • tau2-bench: has not measured this model.
  • Berkeley Function Calling Leaderboard (BFCL) V4: pending. Its latest results are dated 16 Dec 2025, before this model was released.
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 21st of 51 available · 12th of 23 at its price
BenchmarkGPT-5.4Rank among availableBest available
APEX-AgentsAPEX-Agents: tasks passed · reported 52.4% 19th of 37 Claude Sonnet 5.5 75.5%
  • GDP.pdf: has not measured this model.
  • Remote Labor Index: has not measured this model.
Reasoning and knowledgeHard questions across science, maths and general knowledge 32nd of 187 available · 19th of 39 at its price
BenchmarkGPT-5.4Rank among availableBest available
Artificial AnalysisArtificial Analysis: Artificial Analysis Intelligence Index · reported 39.0 31st of 185 Claude Opus 5.5 57.6
Artificial AnalysisArtificial Analysis: Humanity's Last Exam · reported 43.7% 28th of 186 Claude Opus 5.5 61.4%
Artificial AnalysisArtificial Analysis: GPQA Diamond · reported 92.0% 22nd of 168 GPT-6 Astra 96.3%
WritingWhat people prefer in blind comparisons, and judged writing quality 21st of 169 available · 11th of 34 at its price
BenchmarkGPT-5.4Rank among availableBest available
Arena (formerly LMArena)Arena Text: overall · reported 1,475 26th of 153 Gemini 4 Argon 1,525
UGI LeaderboardUGI: writing score · reported 68.8 13th of 124 Gemini 3.8 Flash 78.6
Sticking to the factsSummarising without inventing things, and factual answers 37th of 139 available · 11th of 31 at its price
BenchmarkGPT-5.4Rank among availableBest available
Vectara Hallucination LeaderboardVectara: hallucination rate · reported 7.0% 22nd of 71 GPT-5.4 nano 3.1%
Arena (formerly LMArena)Arena Text factuality: overall · reported 1,485 7th of 109 Gemini 4 Argon 1,511
SimpleQA Verified (Epoch AI)SimpleQA Verified: correct answers · reported 45.1% 34th of 68 GPT-6 Astra 75.6%
Speed under pressureGood decisions against a real clock (fast chess) Not measured
  • BulletBench: has not measured this model.

Nearest alternatives

Against GPT-5.4 at $5.62 blended, Intelligence Index 39.0. "Similar intelligence" means within 4 points or better.

Where to run it

5 hosts on OpenRouter at the standard service tier, with a median of $6.19 blended per million tokens. Cheapest first.

HostQuantisationInputOutputBlended
Azurenot stated$2.50$15.00$5.62
OpenAInot stated$2.50$15.00$5.62
Amazon Bedrocknot stated$2.75$16.50$6.19
Azurenot stated$2.75$16.50$6.19
Azurenot stated$2.75$16.50$6.19

Other service tiers, not counted above: OpenAI fast $5.00 in, $30.00 out; OpenAI flex $1.25 in, $7.50 out. Flex and batch are cheaper but slower or queued; priority and fast cost more.

Prices per million tokens from OpenRouter's public endpoint list, checked 10 Nov 2026.

Against the alternatives

The models you would most likely weigh it against: the leaders of the groups above.

Scroll sideways to see every alternative.

GPT-5.4Claude Opus 5.5
most intelligent at $3–10 per million tokens
MiMo-V2.6-Flash
cheapest with similar intelligence
Claude Fable 5.2
leads overall
GPT-6.1 Sol
OpenAI's best other model
Price per million tokens $5.62 $8.00$0.18$16.00$4.00
Tokens a second – 9646–53
Intelligence Index 39.0 57.637.9–51.8
Spring Prompt overall – 82–9080
Rank among available models, by use case
Product listings – 13th–5th2nd
Decks from an analysis – 7th–4th5th
User surveys – 4th–1st3rd
Marketing planning – 9th–1st–
Agents and tool use 26th of 159 3rd42nd–6th
Professional work 21st of 51 3rd––9th
Reasoning and knowledge 32nd of 187 1st55th–6th
Writing 21st of 169 5th56th–16th
Sticking to the facts 37th of 139 3rd46th–18th
Speed under pressure – ––8th3rd

Shaded figures are better than GPT-5.4's. A dash means no result.

Research and examples

Where it has been measured

16th most intelligent of the 38 models priced $3–10 per million tokens. Among models available today, in the top quarter (by percentile) for writing, reasoning and knowledge, and agents and tool use.

Artificial Analysis lists it but has not published a speed for it yet

Every result

90 published results from 10 sources, each in the source's own units, with the configuration that produced it.

APEX-Agents · reported by APEX-Agents · 1 result
MeasureValueRankConfigurationDated
APEX-Agents: tasks passed% of tasks, higher is better 52.4% 20 of 39 GPT-5.4 1 Oct 2026

Not shown: Not your firm's documents or tools; graded by rubric, not by a client.

Arena (formerly LMArena) · reported by Arena (formerly LMArena) · 20 results
MeasureValueRankConfigurationDated
Arena Agent: confirmed task successIPS effect estimate, higher is better -0.03
-0.05–-0.00
34 of 50 GPT-5.4 (high reasoning, arena agent) 2 Oct 2026
Arena Agent: praise over complaintIPS effect estimate, higher is better 0.00
-0.03–0.04
27 of 50 GPT-5.4 (high reasoning, arena agent) 2 Oct 2026
Arena Agent: steerabilityIPS effect estimate, higher is better 0.05
0.04–0.06
13 of 50 GPT-5.4 (high reasoning, arena agent) 2 Oct 2026
Arena Agent: tool groundingIPS effect estimate, higher is better 0.00
-0.00–0.00
30 of 50 GPT-5.4 (high reasoning, arena agent) 2 Oct 2026
Arena Document: overallArena rating, higher is better 1,470
1,464–1,476
18 of 44 GPT-5.4 13 Sep 2026
Arena Search: overallArena rating, higher is better 1,195
1,189–1,200
14 of 34 GPT-5.4 (arena search) 24 Aug 2026
Arena Text factuality: overallArena rating, higher is better 1,475
1,472–1,478
22 of 182 GPT-5.4 2 Oct 2026
Arena Text factuality: overallArena rating, higher is better 1,485
1,482–1,488
8 of 182 GPT-5.4 (high reasoning) 2 Oct 2026
Arena Text: business, management and financeArena rating, higher is better 1,474
1,468–1,481
33 of 406 GPT-5.4 2 Oct 2026
Arena Text: business, management and financeArena rating, higher is better 1,482
1,475–1,488
25 of 406 GPT-5.4 (high reasoning) 2 Oct 2026
Arena Text: creative writingArena rating, higher is better 1,435
1,428–1,441
66 of 411 GPT-5.4 2 Oct 2026
Arena Text: creative writingArena rating, higher is better 1,442
1,436–1,449
55 of 411 GPT-5.4 (high reasoning) 2 Oct 2026
Arena Text: expert promptsArena rating, higher is better 1,492
1,484–1,500
56 of 363 GPT-5.4 2 Oct 2026
Arena Text: expert promptsArena rating, higher is better 1,515
1,507–1,524
23 of 363 GPT-5.4 (high reasoning) 2 Oct 2026
Arena Text: instruction followingArena rating, higher is better 1,460
1,455–1,465
51 of 413 GPT-5.4 2 Oct 2026
Arena Text: instruction followingArena rating, higher is better 1,471
1,466–1,477
36 of 413 GPT-5.4 (high reasoning) 2 Oct 2026
Arena Text: overallArena rating, higher is better 1,465
1,461–1,468
53 of 413 GPT-5.4 2 Oct 2026
Arena Text: overallArena rating, higher is better 1,475
1,471–1,479
34 of 413 GPT-5.4 (high reasoning) 2 Oct 2026
Arena Text: writing, literature and languageArena rating, higher is better 1,458
1,452–1,464
41 of 412 GPT-5.4 2 Oct 2026
Arena Text: writing, literature and languageArena rating, higher is better 1,465
1,459–1,471
32 of 412 GPT-5.4 (high reasoning) 2 Oct 2026

Not shown: Not a success rate: the estimated effect of choosing this model as the agent's orchestrator, against an even mix of the models tested, inside Arena's own harness.

Artificial Analysis · reported by Artificial Analysis · 24 results
MeasureValueRankConfigurationDated
Artificial Analysis: Artificial Analysis Coding Indexindex score, higher is better 71.1 49 of 187 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 27.6 132 of 413 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 18.2 227 of 413 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: Artificial Analysis Intelligence Indexindex score, higher is better 39.0 70 of 413 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 87.1% 104 of 360 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 74.8% 219 of 360 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: GPQA Diamond% of questions, higher is better 92.0% 42 of 360 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 30.8% 141 of 419 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 11.3% 266 of 419 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: Humanity's Last Exam% of questions, higher is better 43.7% 55 of 419 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 65.9% 85 of 288 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 48.4% 147 of 288 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: IFBench% of instructions, higher is better 73.9% 34 of 288 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 76.7% 137 of 407 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 58.3% 256 of 407 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: Long-context reasoning (AA-LCR)% of questions, higher is better 82.0% 44 of 407 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench 2.1% of tasks, higher is better 78.3% 53 of 188 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 43.2% 40 of 282 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 37.9% 60 of 282 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: Terminal-Bench Hard% of tasks, higher is better 57.6% 11 of 282 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: τ-bench banking% of tasks, higher is better 39.6% 32 of 175 GPT-5.4 (xhigh reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 74.6% 115 of 286 GPT-5.4 (low reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 36.0% 190 of 286 GPT-5.4 (no reasoning) 20 Oct 2026
Artificial Analysis: τ²-bench telecom% of tasks, higher is better 87.1% 65 of 286 GPT-5.4 (xhigh reasoning) 20 Oct 2026

Not shown: Not your codebase or tools.

Microsoft STATE-Bench · reported by Microsoft STATE-Bench · 18 results
MeasureValueRankConfigurationDated
STATE-Bench v0.7: consistency (pass^5)% of tasks, higher is better 26.2% 4 of 5 GPT-5.4 (default reasoning, statebench) 25 May 2026
STATE-Bench v0.7: consistency (pass^5)% of tasks, higher is better 38.0% 1 of 5 GPT-5.4 (high reasoning, statebench) 25 May 2026
STATE-Bench v0.7: customer support task success (pass@1)% of tasks, higher is better 47.2% 3 of 5 GPT-5.4 (default reasoning, statebench) 25 May 2026
STATE-Bench v0.7: customer support task success (pass@1)% of tasks, higher is better 57.6% 1 of 5 GPT-5.4 (high reasoning, statebench) 25 May 2026
STATE-Bench v0.7: shopping assistant task success (pass@1)% of tasks, higher is better 50.3% 3 of 5 GPT-5.4 (default reasoning, statebench) 25 May 2026
STATE-Bench v0.7: shopping assistant task success (pass@1)% of tasks, higher is better 53.6% 2 of 5 GPT-5.4 (high reasoning, statebench) 25 May 2026
STATE-Bench v0.7: task success (pass@1)% of tasks, higher is better 46.9% 5 of 5 GPT-5.4 (default reasoning, statebench) 25 May 2026
STATE-Bench v0.7: task success (pass@1)% of tasks, higher is better 55.7% 1 of 5 GPT-5.4 (high reasoning, statebench) 25 May 2026
STATE-Bench v0.7: travel task success (pass@1)% of tasks, higher is better 43.2% 5 of 5 GPT-5.4 (default reasoning, statebench) 25 May 2026
STATE-Bench v0.7: travel task success (pass@1)% of tasks, higher is better 55.9% 1 of 5 GPT-5.4 (high reasoning, statebench) 25 May 2026
STATE-Bench v0.7: user experiencescore from 1 to 5, higher is better 3.41 3 of 5 GPT-5.4 (default reasoning, statebench) 25 May 2026
STATE-Bench v0.7: user experiencescore from 1 to 5, higher is better 3.49 1 of 5 GPT-5.4 (high reasoning, statebench) 25 May 2026
STATE-Bench v0.8: consistency (pass^5)% of tasks, higher is better 41.8% 2 of 2 GPT-5.4 (high reasoning, statebench) 1 Jul 2026
STATE-Bench v0.8: customer support task success (pass@1)% of tasks, higher is better 59.0% 1 of 2 GPT-5.4 (high reasoning, statebench) 1 Jul 2026
STATE-Bench v0.8: shopping assistant task success (pass@1)% of tasks, higher is better 55.0% 1 of 2 GPT-5.4 (high reasoning, statebench) 1 Jul 2026
STATE-Bench v0.8: task success (pass@1)% of tasks, higher is better 58.6% 2 of 2 GPT-5.4 (high reasoning, statebench) 1 Jul 2026
STATE-Bench v0.8: travel task success (pass@1)% of tasks, higher is better 62.0% 1 of 2 GPT-5.4 (high reasoning, statebench) 1 Jul 2026
STATE-Bench v0.8: user experiencescore from 1 to 5, higher is better 3.67 1 of 2 GPT-5.4 (high reasoning, statebench) 1 Jul 2026

Not shown: Not the model alone: it runs inside STATE-Bench's agent loop against simulated users, and versions of the benchmark are not comparable.

OpenHands Index · reported by OpenHands Index · 6 results
MeasureValueRankConfigurationDated
OpenHands Index: average% resolved, higher is better 64.3% 6 of 34 GPT-5.4 (openhands) 30 Jun 2026
OpenHands Index: front end (SWE-Bench Multimodal)% resolved, higher is better 36.8% 11 of 34 GPT-5.4 (openhands) 30 Jun 2026
OpenHands Index: greenfield (Commit0)% resolved, higher is better 56.2% 3 of 34 GPT-5.4 (openhands) 30 Jun 2026
OpenHands Index: information gathering (GAIA)% resolved, higher is better 82.4% 3 of 34 GPT-5.4 (openhands) 30 Jun 2026
OpenHands Index: issue resolution (SWE-Bench)% resolved, higher is better 75.6% 10 of 34 GPT-5.4 (openhands) 30 Jun 2026
OpenHands Index: testing (SWT-Bench)% resolved, higher is better 70.4% 12 of 34 GPT-5.4 (openhands) 30 Jun 2026

Not shown: Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.

SimpleQA Verified (Epoch AI) · reported by SimpleQA Verified (Epoch AI) · 1 result
MeasureValueRankConfigurationDated
SimpleQA Verified: correct answers% of questions, higher is better 45.1% 38 of 85 GPT-5.4 (xhigh reasoning) 27 Aug 2026

Not shown: Not answers grounded in your documents; tests what the model remembers.

tau2-bench · reported by tau2-bench · 2 results
MeasureValueRankConfigurationDated
tau2 v1.0.1, banking knowledge, alltools retrieval: consistency (pass^4)% of tasks, higher is better 21.7% 9 of 21 GPT-5.4 (xhigh reasoning, tau2) 6 May 2026
tau2 v1.0.1, banking knowledge, alltools retrieval: task success (pass^1)% of tasks, higher is better 39.4% 10 of 21 GPT-5.4 (xhigh reasoning, tau2) 6 May 2026

Not shown: Not your policies or customers: the tasks and policies are tau2's own, and the customer is simulated by gpt-5.2.

UGI Leaderboard · reported by UGI Leaderboard · 15 results
MeasureValueRankConfigurationDated
UGI: requested-length error% off the requested word count, lower is better 46.0% 317 of 379 GPT-5.4 (high reasoning) 5 Mar 2026
UGI: requested-length error% off the requested word count, lower is better 57.0% 327 of 379 GPT-5.4 (low reasoning) 5 Mar 2026
UGI: requested-length error% off the requested word count, lower is better 53.0% 323 of 379 GPT-5.4 (medium reasoning) 5 Mar 2026
UGI: requested-length error% off the requested word count, lower is better 50.0% 321 of 379 GPT-5.4 (no reasoning) 5 Mar 2026
UGI: requested-length error% off the requested word count, lower is better 30.0% 262 of 379 GPT-5.4 (xhigh reasoning) 5 Mar 2026
UGI: style adherencescore from 0 to 1, higher is better 0.40 27 of 379 GPT-5.4 (high reasoning) 5 Mar 2026
UGI: style adherencescore from 0 to 1, higher is better 0.38 65 of 379 GPT-5.4 (low reasoning) 5 Mar 2026
UGI: style adherencescore from 0 to 1, higher is better 0.39 35 of 379 GPT-5.4 (medium reasoning) 5 Mar 2026
UGI: style adherencescore from 0 to 1, higher is better 0.38 61 of 379 GPT-5.4 (no reasoning) 5 Mar 2026
UGI: style adherencescore from 0 to 1, higher is better 0.40 30 of 379 GPT-5.4 (xhigh reasoning) 5 Mar 2026
UGI: writing scorescore out of 100, higher is better 68.8 33 of 379 GPT-5.4 (high reasoning) 5 Mar 2026
UGI: writing scorescore out of 100, higher is better 64.9 68 of 379 GPT-5.4 (low reasoning) 5 Mar 2026
UGI: writing scorescore out of 100, higher is better 67.3 47 of 379 GPT-5.4 (medium reasoning) 5 Mar 2026
UGI: writing scorescore out of 100, higher is better 58.0 116 of 379 GPT-5.4 (no reasoning) 5 Mar 2026
UGI: writing scorescore out of 100, higher is better 68.7 34 of 379 GPT-5.4 (xhigh reasoning) 5 Mar 2026

Not shown: Not other format limits such as character counts or bullet counts.

Vectara Hallucination Leaderboard · reported by Vectara Hallucination Leaderboard · 2 results
MeasureValueRankConfigurationDated
Vectara: answer rate% of documents, higher is better 99.9% 19 of 108 GPT-5.4 22 Sep 2026
Vectara: hallucination rate% of summaries, lower is better 7.0% 32 of 108 GPT-5.4 22 Sep 2026

Not shown: Not a quality score: a low rate usually means content filters were triggered, and hallucination rates are measured on answered documents only.

Vending-Bench 2 · reported by Vending-Bench 2 · 1 result
MeasureValueRankConfigurationDated
Vending-Bench 2: money after a yearUS dollars, higher is better $6,144 16 of 63 GPT-5.4 1 Oct 2026

Not shown: Not a real business; one simulated market with set rules.

Will GPT-5.4 work for your product?

Benchmarks aren't your task. Paste the prompt your product uses and run it beside the models you could use instead, with every answer checked and costed.

New here? Your email starts a free trial. No credit card needed.