GPT-5.4 vs GPT-6 Astra

GPT-6 Astra leads GPT-5.4 on coding, 85.1 to 64.4 on the Vibe Coding Index, and GPT-5.4 costs 3.3x less per output token ($15.00 vs $50.00 per million). Updated October 22, 2026; 83 benchmarks have a result for both.

Replace GPT-5.4 Replace GPT-6 Astra

GPT-5.4OpenAI

Coding: 64.4
Overall: 69.6
Output $/M: $15.00

GPT-6 AstraOpenAI

Coding: 85.1
Overall: 83.0
Output $/M: $50.00

vs boards won of 83 shared

GPT-5.4 or GPT-6 Astra: which to pick

Pick GPT-6 Astra for coding: it scores 85.1 on the Vibe Coding Index against 64.4 for GPT-5.4.

Pick GPT-5.4 for output tokens 3.3x cheaper ($15.00 vs $50.00 per million) and work measured by LMArena Text: coding or LMArena Text: software and IT, where it leads.

Model AGPT-5.4OpenAI64.4/1003 independent benchmarks$1.02 per solved task Model BGPT-6 AstraOpenAI85.1/1005 independent benchmarksNo cost per solved task

832738

Shared results: 83
GPT-5.4 only: 27
GPT-6 Astra only: 38
Like-for-like categories: 8 of 9

Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

Coding
Vibe Coding Index, six public coding benchmarks
GPT-6 Astra
GPT-6 Astra scores 85.1 against 64.4.
3 and 5 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-6 Astra
GPT-6 Astra leads, 52.7 against 39.0.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only GPT-6 Astra has a AA Agentic Index result, 51.0.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5.4 has a GPQA Diamond result, 91.7%.
Directional only
Math
AIME
No call
Only GPT-5.4 has a AIME result, 96.7%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.4 has a MMLU Pro result, 87.5%.
Directional only
Web design
Design Arena: websites
GPT-6 Astra
GPT-6 Astra leads, 1336 against 1198.
Same board, both measured
General
Vals Index
No call
Only GPT-6 Astra has a Vals Index result, 65.0%.
Directional only
Long documents
Listed context window
No call
Both take 1.05M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GPT-5.4
GPT-5.4 $0.010, GPT-6 Astra $0.035 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.4 has the number, $1.02.
Directional only

At a glance

Coding index: GPT-5.4 64.4 from 3 benchmarks, GPT-6 Astra 85.1 from 5 benchmarks.
Widest category: Agentic work, 34.6 points apart on average over 6 shared benchmarks.
Widest single gap: Vending-Bench 2: GPT-5.4 $6,144 against GPT-6 Astra $15,515, 58.6 points apart on a 0 to 100 scale.
Output price: GPT-5.4 $15.00 against GPT-6 Astra $50.00 per million output tokens, 3.3x apart.
Context window: Both take 1.05M tokens.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
Category	GPT-5.4	GPT-6 Astra
Coding	2	12
Intelligence	0	2
Agentic work	0	6
Reasoning	4	9
Math	2	4
Knowledge	4	2
Web design	0	8
General	21	7
All shared boards	33	50

CodingLike-for-likeSix of these boards feed the Vibe Coding Index212

Coding benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
LiveBench coding	Not measured	93.7%	Only GPT-6 Astra measured
LMArena WebDev	1464	1774	GPT-6 Astra
AA Coding Index	71.1	76.9	GPT-6 Astra
LiveCodeBench (Vals AI)	84.1%	Not measured	Only GPT-5.4 measured
SWE-bench Verified (Vals AI)	78.2%	Not measured	Only GPT-5.4 measured
Terminal-Bench 4 (Vals AI)	Not measured	59.6%	Only GPT-6 Astra measured
Vibe Code Bench (Vals AI)	67.4%	89.6%	GPT-6 Astra
Code migration (Vals AI)	35.0%	67.7%	GPT-6 Astra
ProgramBench (Vals AI)	0.5%	5.5%	GPT-6 Astra
IOI (Vals AI)	Not measured	100.0%	Only GPT-6 Astra measured
Terminal-Bench 2.1 (Vals AI)	Not measured	87.3%	Only GPT-6 Astra measured
SWE-bench Multimodal (Vals AI)	Not measured	43.3%	Only GPT-6 Astra measured
Vibe Code Bench 1-100 (Vals AI)	Not measured	27.6%	Only GPT-6 Astra measured
SkillsBench (Vals AI)	51.7%	Not measured	Only GPT-5.4 measured
CyberBench (Vals AI)	Not measured	41.1%	Only GPT-6 Astra measured
SciCode	Not measured	56.5%	Only GPT-6 Astra measured
Terminal-Bench Hard (Artificial Analysis run)	57.6%	Not measured	Only GPT-5.4 measured
Terminal-Bench	81.8%	Not measured	Only GPT-5.4 measured
FrontierCode	Not measured	53.3%	Only GPT-6 Astra measured
FrontierSWE	Not measured	65.5%	Only GPT-6 Astra measured
GSO-Bench	31.4%	79.4%	GPT-6 Astra
WeirdML	77.7%	93.3%	GPT-6 Astra
MirrorCode	15.6%	46.7%	GPT-6 Astra
ALE-Bench	1607.0	2951.3	GPT-6 Astra
LMArena Text: coding	1496	1489	GPT-5.4
LMArena Text: software and IT	1491	1474	GPT-5.4
LMArena WebDev: React	1454	1798	GPT-6 Astra
LMArena WebDev: HTML	1478	1775	GPT-6 Astra
LMArena WebDev: from an image	1436	1726	GPT-6 Astra
SciCode	Not measured	12.3%	Only GPT-6 Astra measured
Neither model has a result on the other 9 boards here

IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02

Intelligence benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
AA Intelligence Index	39.0	52.7	GPT-6 Astra
Epoch Capabilities Index	157.3	168.8	GPT-6 Astra

Agentic workLike-for-likePublic boards, shown as published, kept out of the index06

Agentic work benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
AA Agentic Index	Not measured	51.0	Only GPT-6 Astra measured
Design Arena: agents building web apps	1059	1373	GPT-6 Astra
Design Arena: agents building full-stack apps	1017	1357	GPT-6 Astra
Design Arena: agents building mobile apps	1087	1312	GPT-6 Astra
Design Arena: agents building Android apps	1031	Not measured	Only GPT-5.4 measured
Design Arena: agents building Godot games	1135	1348	GPT-6 Astra
Design Arena: PowerPoint slides	Not measured	1219	Only GPT-6 Astra measured
Design Arena: spreadsheets	Not measured	1365	Only GPT-6 Astra measured
LiveBench research	Not measured	85.2%	Only GPT-6 Astra measured
LiveBench browser	Not measured	84.0%	Only GPT-6 Astra measured
LiveBench long agentic	Not measured	82.4%	Only GPT-6 Astra measured
SREBench (Vals AI)	Not measured	56.9%	Only GPT-6 Astra measured
Finance Agent (Vals AI)	Not measured	53.5%	Only GPT-6 Astra measured
Tax Agent Bench (Vals AI)	Not measured	20.7%	Only GPT-6 Astra measured
Legal Research Bench (Vals AI)	Not measured	39.4%	Only GPT-6 Astra measured
Harvey's Legal Agent Benchmark (Vals AI)	Not measured	7.1%	Only GPT-6 Astra measured
Excel Modeling Benchmark (Vals AI)	Not measured	71.7%	Only GPT-6 Astra measured
BioMysteryBench (Vals AI)	Not measured	79.3%	Only GPT-6 Astra measured
Terminal-Bench-Science (Vals AI)	Not measured	62.9%	Only GPT-6 Astra measured
Vals Science Index	Not measured	65.1%	Only GPT-6 Astra measured
Time Horizon Index (Vals AI)	Not measured	90.5%	Only GPT-6 Astra measured
CUA-bench (Vals AI)	Not measured	19.2%	Only GPT-6 Astra measured
Robo-bench (Vals AI)	Not measured	23.3%	Only GPT-6 Astra measured
τ²-Bench Telecom	87.1%	Not measured	Only GPT-5.4 measured
APEX-Agents	52.4%	64.7%	GPT-6 Astra
METR time horizons	74.3%	Not measured	Only GPT-5.4 measured
Remote Labor Index	Not measured	20.8%	Only GPT-6 Astra measured
BALROG	Not measured	68.3%	Only GPT-6 Astra measured
DeepResearch Bench	35.1%	Not measured	Only GPT-5.4 measured
PostTrainBench	19.0%	Not measured	Only GPT-5.4 measured
SpireClimb	Not measured	25.7%	Only GPT-6 Astra measured
Vending-Bench 2	$6,144	$15,515	GPT-6 Astra
Neither model has a result on the other 9 boards here

ReasoningLike-for-likePublic boards, shown as published, kept out of the index49

Reasoning benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
GPQA Diamond	91.7%	Not measured	Only GPT-5.4 measured
LiveBench reasoning	Not measured	92.7%	Only GPT-6 Astra measured
MysteryMechanism (Vals AI)	Not measured	53.1%	Only GPT-6 Astra measured
MMMU Pro (Vals AI)	87.5%	Not measured	Only GPT-5.4 measured
GPQA Diamond (Artificial Analysis run)	92.0%	96.3%	GPT-6 Astra
Humanity's Last Exam (Artificial Analysis run)	43.7%	54.7%	GPT-6 Astra
GPQA Diamond (Epoch AI run)	93.3%	95.8%	GPT-6 Astra
ARC-AGI	93.7%	98.5%	GPT-6 Astra
ARC-AGI-2	74.0%	95.0%	GPT-6 Astra
Chess puzzles	48.0%	74.0%	GPT-6 Astra
Mystery game puzzles	37.0%	84.0%	GPT-6 Astra
EBR-bench	25.4%	76.2%	GPT-6 Astra
LMArena Text: hard prompts	1485	1463	GPT-5.4
LMArena Text: expert	1507	1489	GPT-5.4
LMArena Text: hard prompts in English	1479	1464	GPT-5.4
LMArena Vision: diagrams	1327	1297	GPT-5.4
LMArena Vision: homework	1336	1337	GPT-6 Astra
Kagi LLM Benchmark	63.8%	Not measured	Only GPT-5.4 measured
Neither model has a result on the other 4 boards here

MathLike-for-likePublic boards, shown as published, kept out of the index24

Math benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
AIME	96.7%	Not measured	Only GPT-5.4 measured
OTIS Mock AIME	97.8%	100.0%	GPT-6 Astra
FrontierMath, tiers 1 to 3	78.6%	93.7%	GPT-6 Astra
FrontierMath, tier 4	49.0%	97.6%	GPT-6 Astra
ProofBench	56.0%	99.0%	GPT-6 Astra
LMArena Text: math	1488	1465	GPT-5.4
LMArena Text: mathematical work	1497	1484	GPT-5.4
Neither model has a result on the other 6 boards here

KnowledgeLike-for-likePublic boards, shown as published, kept out of the index42

Knowledge benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
MMLU Pro	87.5%	Not measured	Only GPT-5.4 measured
LegalBench (Vals AI)	86.0%	Not measured	Only GPT-5.4 measured
CaseLaw (Vals AI)	63.8%	Not measured	Only GPT-5.4 measured
CorpFin (Vals AI)	65.3%	Not measured	Only GPT-5.4 measured
TaxEval (Vals AI)	74.0%	Not measured	Only GPT-5.4 measured
MortgageTax (Vals AI)	68.3%	Not measured	Only GPT-5.4 measured
MedQA (Vals AI)	96.1%	Not measured	Only GPT-5.4 measured
MedCode (Vals AI)	41.3%	48.5%	GPT-6 Astra
SimpleQA Verified	45.1%	75.6%	GPT-6 Astra
LMArena Text: science	1479	1451	GPT-5.4
LMArena Text: medicine and healthcare	1458	1425	GPT-5.4
LMArena Text: legal and government	1486	1462	GPT-5.4
LMArena Text: business and finance	1474	1439	GPT-5.4
FACTS Benchmark Suite	Not measured	75.7%	Only GPT-6 Astra measured
Neither model has a result on the other 4 boards here

Web designLike-for-likePublic boards, shown as published, kept out of the index08

Web design benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
Design Arena: websites	1198	1336	GPT-6 Astra
Design Arena: UI components	1212	1393	GPT-6 Astra
Design Arena: game dev	1194	1478	GPT-6 Astra
Design Arena: data viz	1196	1335	GPT-6 Astra
Design Arena: 3D	1065	1500	GPT-6 Astra
Design Arena: all code categories	1179	1379	GPT-6 Astra
Design Arena: SVG	1141	1431	GPT-6 Astra
Design Arena: ASCII art	1188	1353	GPT-6 Astra
LiveBench design	Not measured	82.0%	Only GPT-6 Astra measured

GeneralLike-for-likePublic boards, shown as published, kept out of the index217

General benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
Vals Index	Not measured	65.0%	Only GPT-6 Astra measured
LiveBench data analysis	Not measured	83.0%	Only GPT-6 Astra measured
SAGE (Vals AI)	43.3%	46.4%	GPT-6 Astra
MedScribe (Vals AI)	77.5%	87.9%	GPT-6 Astra
IFBench	73.9%	Not measured	Only GPT-5.4 measured
Long context reasoning (AA-LCR)	82.0%	80.7%	GPT-5.4
DTBench	94.4%	97.3%	GPT-6 Astra
LMCA	52.0%	64.4%	GPT-6 Astra
CL-bench	27.9%	Not measured	Only GPT-5.4 measured
Furniture assembly	37.5%	80.0%	GPT-6 Astra
CL-bench Life	21.7%	Not measured	Only GPT-5.4 measured
LMArena Text	1469	1441	GPT-5.4
LMArena Text: instruction following	1469	1455	GPT-5.4
LMArena Text: creative writing	1439	1420	GPT-5.4
LMArena Vision	1303	1280	GPT-5.4
LMArena Text: English	1467	1447	GPT-5.4
LMArena Text: non-English	1465	1432	GPT-5.4
LMArena Text: Chinese	1520	1489	GPT-5.4
LMArena Text: German	1472	1459	GPT-5.4
LMArena Text: Japanese	1485	1405	GPT-5.4
LMArena Text: Korean	1449	1428	GPT-5.4
LMArena Text: French	1493	1432	GPT-5.4
LMArena Text: Spanish	1452	1422	GPT-5.4
LMArena Text: Russian	1480	1438	GPT-5.4
LMArena Text: Polish	1472	1422	GPT-5.4
LMArena Text: multi-turn	1482	1447	GPT-5.4
LMArena Text: longer queries	1473	1460	GPT-5.4
LMArena Text: writing and language	1458	1437	GPT-5.4
LMArena Text: entertainment and media	1438	1427	GPT-5.4
LMArena Vision: English	1301	1280	GPT-5.4
LMArena Vision: Chinese	1354	1361	GPT-6 Astra
LMArena Vision: reading text	1312	1292	GPT-5.4
LMArena Vision: humor	1284	Not measured	Only GPT-5.4 measured
LMArena Vision: recognizing things	1251	Not measured	Only GPT-5.4 measured
LMArena Vision: creative writing	1283	1286	GPT-6 Astra
LMArena Document	1471	Not measured	Only GPT-5.4 measured
Neither model has a result on the other 10 boards here

Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared

Image generation benchmarks for GPT-5.4 and GPT-6 Astra
Benchmark	GPT-5.4	GPT-6 Astra	Leader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

GPT-5.4$0.010Fits in one request

GPT-6 Astra$0.035Fits in one request

GPT-5.4 costs less

Repository review

50K input + 3K output tokens

GPT-5.4$0.170Fits in one request

GPT-6 Astra$0.650Fits in one request

GPT-5.4 costs less

Agent loop

200K input + 10K output tokens

GPT-5.4$0.650Fits in one request

GPT-6 Astra$2.50Fits in one request

GPT-5.4 costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of GPT-5.4 and GPT-6 Astra
Specification	OpenAI GPT-5.4	OpenAI GPT-6 Astra
Maker	OpenAI	OpenAI
Weights	No public weights listed	No public weights listed
Released	March 5, 2026	September 4, 2026
Context window	1.05M	1.05M
Input $/Mtok	$2.50	$10.00
Output $/Mtok	$15.00	$50.00
Coding index	64.4	85.1
Independent measurements behind it	3	5
Cost per solved task	$1.02	Not computable
Source	OpenRouter listing	OpenRouter listing

Questions

Is GPT-5.4 better than GPT-6 Astra for coding?

Pick GPT-6 Astra for coding: it scores 85.1 on the Vibe Coding Index against 64.4 for GPT-5.4. Pick GPT-5.4 for output tokens 3.3x cheaper ($15.00 vs $50.00 per million) and work measured by LMArena Text: coding or LMArena Text: software and IT, where it leads.

Which is cheaper, GPT-5.4 or GPT-6 Astra?

GPT-5.4 costs $2.50 per million input tokens and $15.00 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. Cost per solved task is known for GPT-5.4 only.

Which has the bigger context window, GPT-5.4 or GPT-6 Astra?

Neither: both take 1.05M tokens.

How many benchmarks is the GPT-5.4 vs GPT-6 Astra comparison based on?

83 benchmarks have a result for both models as of October 22, 2026. GPT-5.4's coding index rests on 3 independent benchmarks and GPT-6 Astra's on 5. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/gpt-5-4-vs-gpt-6-astra" title="GPT-5.4 vs GPT-6 Astra, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/gpt-5-4-vs-gpt-6-astra">LLMBenchmark.ai, GPT-5.4 vs GPT-6 Astra</a></p>

GPT-5.4 or GPT-6 Astra: which to pick

Which one for your work

Coding

Intelligence

Agentic work

Reasoning

Math

Knowledge

Web design

General

Long documents

Everyday cost

Cost per solved task

At a glance

Benchmark by benchmark

What each workload costs

Chat turn

Repository review

Agent loop

Specification differences

Questions

Compare next