Claude Opus 4.8 vs GPT-5.4 Mini

Claude Opus 4.8 leads GPT-5.4 Mini on coding, 65.0 to 35.8 on the Vibe Coding Index, and GPT-5.4 Mini costs 5.6x less per output token ($4.50 vs $25.00 per million). Updated October 21, 2026; 85 benchmarks have a result for both.

Replace Claude Opus 4.8 Replace GPT-5.4 Mini

Claude Opus 4.8Anthropic

Coding: 65.0
Overall: 69.8
Output $/M: $25.00

GPT-5.4 MiniOpenAI

Coding: 35.8
Overall: 44.0
Output $/M: $4.50

vs boards won of 85 shared

Claude Opus 4.8 or GPT-5.4 Mini: which to pick

Pick Claude Opus 4.8 for coding: it scores 65.0 on the Vibe Coding Index against 35.8 for GPT-5.4 Mini.

Pick GPT-5.4 Mini for output tokens 5.6x cheaper ($4.50 vs $25.00 per million) and a lower cost per solved task ($0.70 against $2.17).

Model AClaude Opus 4.8Anthropic65.0/1006 independent benchmarks$2.17 per solved task Model BGPT-5.4 MiniOpenAI35.8/1005 independent benchmarks$0.70 per solved task

85454

Shared results: 85
Claude Opus 4.8 only: 45
GPT-5.4 Mini only: 4
Like-for-like categories: 7 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
Claude Opus 4.8
Claude Opus 4.8 scores 65.0 against 35.8.
6 and 5 benchmarks behind the scores
Intelligence
AA Intelligence Index
Claude Opus 4.8
Claude Opus 4.8 leads, 41.8 against 24.1.
Same board, both measured
Agentic work
AA Agentic Index
Claude Opus 4.8
Claude Opus 4.8 leads, 41.9 against 17.9.
Same board, both measured
Reasoning
GPQA Diamond
Claude Opus 4.8
Claude Opus 4.8 leads, 92.4% against 83.1%.
Same board, both measured
Math
AIME
No call
Only GPT-5.4 Mini has a AIME result, 95.6%.
Directional only
Knowledge
MMLU Pro
Claude Opus 4.8
Claude Opus 4.8 leads, 89.6% against 84.5%.
Same board, both measured
Web design
Design Arena: websites
No call
Only Claude Opus 4.8 has a Design Arena: websites result, 1259.
Directional only
Long documents
Listed context window
Claude Opus 4.8
Claude Opus 4.8 takes 1M tokens against 400K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GPT-5.4 Mini
Claude Opus 4.8 $0.018, GPT-5.4 Mini $0.0030 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
GPT-5.4 Mini
Claude Opus 4.8 $2.17, GPT-5.4 Mini $0.70.
Published attempt cost and pass rate

At a glance

Coding index: Claude Opus 4.8 65.0 from 6 benchmarks, GPT-5.4 Mini 35.8 from 5 benchmarks.
Widest category: Math, 25.9 points apart on average over 6 shared benchmarks.
Widest single gap: ARC-AGI-2: Claude Opus 4.8 72.1% against GPT-5.4 Mini 18.9%, 53.2 points apart on a 0 to 100 scale.
Output price: Claude Opus 4.8 $25.00 against GPT-5.4 Mini $4.50 per million output tokens, 5.6x apart.
Context window: Claude Opus 4.8 takes 1M tokens, GPT-5.4 Mini 400K.

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	Claude Opus 4.8	GPT-5.4 Mini
Coding	17	0
Intelligence	2	0
Agentic work	7	0
Reasoning	15	0
Math	6	0
Knowledge	9	0
General	28	1
All shared boards	84	1

CodingLike-for-likeSix of these boards feed the Vibe Coding Index170

Coding benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
LiveBench coding	77.3%	Not measured	Only Claude Opus 4.8 measured
LMArena WebDev	1556	1397	Claude Opus 4.8
AA Coding Index	74.3	56.1	Claude Opus 4.8
LiveCodeBench (Vals AI)	87.8%	81.5%	Claude Opus 4.8
SWE-bench Verified (Vals AI)	88.6%	73.0%	Claude Opus 4.8
Terminal-Bench 4 (Vals AI)	23.2%	2.5%	Claude Opus 4.8
Vibe Code Bench (Vals AI)	82.7%	48.0%	Claude Opus 4.8
Code migration (Vals AI)	47.3%	12.9%	Claude Opus 4.8
ProgramBench (Vals AI)	1.0%	0.0%	Claude Opus 4.8
Terminal-Bench 2.1 (Vals AI)	71.9%	54.7%	Claude Opus 4.8
SkillsBench (Vals AI)	59.2%	Not measured	Only Claude Opus 4.8 measured
SciCode	54.4%	52.1%	Claude Opus 4.8
Terminal-Bench Hard (Artificial Analysis run)	58.3%	52.3%	Claude Opus 4.8
FrontierCode	46.5%	27.0%	Claude Opus 4.8
GSO-Bench	47.1%	Not measured	Only Claude Opus 4.8 measured
WeirdML	82.9%	60.3%	Claude Opus 4.8
ALE-Bench	Not measured	1188.6	Only GPT-5.4 Mini measured
LMArena Text: coding	1490	1438	Claude Opus 4.8
LMArena Text: software and IT	1488	1436	Claude Opus 4.8
LMArena WebDev: React	1557	1386	Claude Opus 4.8
LMArena WebDev: HTML	1543	1419	Claude Opus 4.8
SciCode	12.3%	Not measured	Only Claude Opus 4.8 measured
Neither model has a result on the other 17 boards here

IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20

Intelligence benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
AA Intelligence Index	41.8	24.1	Claude Opus 4.8
Epoch Capabilities Index	159.0	148.7	Claude Opus 4.8

Agentic workLike-for-likePublic boards, shown as published, kept out of the index70

Agentic work benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
AA Agentic Index	41.9	17.9	Claude Opus 4.8
Design Arena: agents building web apps	1233	Not measured	Only Claude Opus 4.8 measured
Design Arena: agents building full-stack apps	1249	Not measured	Only Claude Opus 4.8 measured
Design Arena: agents building mobile apps	1222	Not measured	Only Claude Opus 4.8 measured
Design Arena: agents building Android apps	1232	Not measured	Only Claude Opus 4.8 measured
Design Arena: agents building Godot games	1231	Not measured	Only Claude Opus 4.8 measured
Design Arena: agents building games	1238	Not measured	Only Claude Opus 4.8 measured
Design Arena: HTML slides	1221	Not measured	Only Claude Opus 4.8 measured
Design Arena: PowerPoint slides	1227	Not measured	Only Claude Opus 4.8 measured
Design Arena: spreadsheets	1270	Not measured	Only Claude Opus 4.8 measured
LiveBench research	81.8%	Not measured	Only Claude Opus 4.8 measured
LiveBench browser	96.0%	Not measured	Only Claude Opus 4.8 measured
LiveBench long agentic	48.3%	Not measured	Only Claude Opus 4.8 measured
Finance Agent (Vals AI)	53.9%	45.4%	Claude Opus 4.8
Tax Agent Bench (Vals AI)	30.0%	8.8%	Claude Opus 4.8
Legal Research Bench (Vals AI)	43.8%	12.5%	Claude Opus 4.8
Excel Modeling Benchmark (Vals AI)	69.4%	45.4%	Claude Opus 4.8
Terminal-Bench-Science (Vals AI)	4.3%	Not measured	Only Claude Opus 4.8 measured
Time Horizon Index (Vals AI)	11.8%	Not measured	Only Claude Opus 4.8 measured
τ²-Bench Telecom	94.4%	83.3%	Claude Opus 4.8
APEX-Agents	48.9%	Not measured	Only Claude Opus 4.8 measured
Remote Labor Index	8.3%	Not measured	Only Claude Opus 4.8 measured
OSWorld 2.0	20.6%	Not measured	Only Claude Opus 4.8 measured
DeepResearch Bench	50.2%	36.3%	Claude Opus 4.8
PostTrainBench	33.8%	Not measured	Only Claude Opus 4.8 measured
Vending-Bench 2	$5,787	Not measured	Only Claude Opus 4.8 measured
LMArena Search	1204	Not measured	Only Claude Opus 4.8 measured
Neither model has a result on the other 14 boards here

ReasoningLike-for-likePublic boards, shown as published, kept out of the index150

Reasoning benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
GPQA Diamond	92.4%	83.1%	Claude Opus 4.8
LiveBench reasoning	89.2%	Not measured	Only Claude Opus 4.8 measured
MMMU Pro (Vals AI)	86.6%	79.3%	Claude Opus 4.8
GPQA Diamond (Artificial Analysis run)	92.0%	87.5%	Claude Opus 4.8
Humanity's Last Exam (Artificial Analysis run)	48.7%	28.1%	Claude Opus 4.8
GPQA Diamond (Epoch AI run)	91.0%	86.9%	Claude Opus 4.8
ARC-AGI	92.5%	63.7%	Claude Opus 4.8
ARC-AGI-2	72.1%	18.9%	Claude Opus 4.8
SimpleBench	64.8%	Not measured	Only Claude Opus 4.8 measured
Chess puzzles	52.0%	23.0%	Claude Opus 4.8
Mystery game puzzles	36.0%	11.0%	Claude Opus 4.8
EBR-bench	28.6%	Not measured	Only Claude Opus 4.8 measured
LMArena Text: hard prompts	1482	1424	Claude Opus 4.8
LMArena Text: expert	1503	1435	Claude Opus 4.8
LMArena Text: hard prompts in English	1487	1420	Claude Opus 4.8
LMArena Vision: diagrams	1315	1266	Claude Opus 4.8
LMArena Vision: homework	1337	1285	Claude Opus 4.8
Kagi LLM Benchmark	88.8%	37.9%	Claude Opus 4.8
Neither model has a result on the other 4 boards here

MathLike-for-likePublic boards, shown as published, kept out of the index60

Math benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
AIME	Not measured	95.6%	Only GPT-5.4 Mini measured
OTIS Mock AIME	98.3%	88.9%	Claude Opus 4.8
FrontierMath, tiers 1 to 3	80.0%	51.2%	Claude Opus 4.8
FrontierMath, tier 4	56.1%	9.8%	Claude Opus 4.8
ProofBench	69.0%	21.0%	Claude Opus 4.8
LMArena Text: math	1486	1418	Claude Opus 4.8
LMArena Text: mathematical work	1497	1429	Claude Opus 4.8
Neither model has a result on the other 6 boards here

KnowledgeLike-for-likePublic boards, shown as published, kept out of the index90

Knowledge benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
MMLU Pro	89.6%	84.5%	Claude Opus 4.8
LegalBench (Vals AI)	83.6%	Not measured	Only Claude Opus 4.8 measured
CaseLaw (Vals AI)	Not measured	51.7%	Only GPT-5.4 Mini measured
CorpFin (Vals AI)	66.7%	60.9%	Claude Opus 4.8
TaxEval (Vals AI)	75.6%	71.2%	Claude Opus 4.8
MortgageTax (Vals AI)	69.9%	63.5%	Claude Opus 4.8
MedCode (Vals AI)	53.2%	Not measured	Only Claude Opus 4.8 measured
SimpleQA Verified	53.0%	29.4%	Claude Opus 4.8
LMArena Text: science	1477	1417	Claude Opus 4.8
LMArena Text: medicine and healthcare	1464	1395	Claude Opus 4.8
LMArena Text: legal and government	1477	1414	Claude Opus 4.8
LMArena Text: business and finance	1458	1418	Claude Opus 4.8
FACTS Benchmark Suite	53.5%	Not measured	Only Claude Opus 4.8 measured
Neither model has a result on the other 5 boards here

Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared

Web design benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
Design Arena: websites	1259	Not measured	Only Claude Opus 4.8 measured
Design Arena: UI components	1257	Not measured	Only Claude Opus 4.8 measured
Design Arena: game dev	1264	Not measured	Only Claude Opus 4.8 measured
Design Arena: data viz	1240	Not measured	Only Claude Opus 4.8 measured
Design Arena: 3D	1223	Not measured	Only Claude Opus 4.8 measured
Design Arena: all code categories	1255	Not measured	Only Claude Opus 4.8 measured
Design Arena: SVG	1194	Not measured	Only Claude Opus 4.8 measured
Design Arena: ASCII art	1277	Not measured	Only Claude Opus 4.8 measured
LiveBench design	78.8%	Not measured	Only Claude Opus 4.8 measured

GeneralLike-for-likePublic boards, shown as published, kept out of the index281

General benchmarks for Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	Leader
LiveBench data analysis	66.0%	Not measured	Only Claude Opus 4.8 measured
Public Benefits Bench (Vals AI)	68.1%	Not measured	Only Claude Opus 4.8 measured
SAGE (Vals AI)	54.8%	50.8%	Claude Opus 4.8
MedScribe (Vals AI)	85.8%	Not measured	Only Claude Opus 4.8 measured
Vals Multimodal Index	70.9%	54.2%	Claude Opus 4.8
IFBench	62.2%	73.3%	GPT-5.4 Mini
Long context reasoning (AA-LCR)	77.7%	77.0%	Claude Opus 4.8
DTBench	94.9%	80.0%	Claude Opus 4.8
LMCA	57.5%	40.8%	Claude Opus 4.8
Surface Evolver Bench	87.5%	Not measured	Only Claude Opus 4.8 measured
Furniture assembly	42.5%	Not measured	Only Claude Opus 4.8 measured
LMArena Text	1461	1412	Claude Opus 4.8
LMArena Text: instruction following	1476	1405	Claude Opus 4.8
LMArena Text: creative writing	1454	1370	Claude Opus 4.8
LMArena Vision	1294	1245	Claude Opus 4.8
LMArena Text: English	1468	1412	Claude Opus 4.8
LMArena Text: non-English	1450	1405	Claude Opus 4.8
LMArena Text: Chinese	1506	1446	Claude Opus 4.8
LMArena Text: German	1472	1408	Claude Opus 4.8
LMArena Text: Japanese	1442	1375	Claude Opus 4.8
LMArena Text: Korean	1431	1370	Claude Opus 4.8
LMArena Text: French	1481	1441	Claude Opus 4.8
LMArena Text: Spanish	1465	1405	Claude Opus 4.8
LMArena Text: Russian	1474	1417	Claude Opus 4.8
LMArena Text: Polish	1459	1415	Claude Opus 4.8
LMArena Text: multi-turn	1476	1429	Claude Opus 4.8
LMArena Text: longer queries	1483	1408	Claude Opus 4.8
LMArena Text: writing and language	1461	1393	Claude Opus 4.8
LMArena Text: entertainment and media	1440	1373	Claude Opus 4.8
LMArena Vision: English	1293	1244	Claude Opus 4.8
LMArena Vision: Chinese	1346	1286	Claude Opus 4.8
LMArena Vision: reading text	1303	1253	Claude Opus 4.8
LMArena Vision: humor	1282	1235	Claude Opus 4.8
LMArena Vision: recognizing things	Not measured	1218	Only GPT-5.4 Mini measured
LMArena Vision: creative writing	1301	1222	Claude Opus 4.8
LMArena Document	1475	Not measured	Only Claude Opus 4.8 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 Claude Opus 4.8 and GPT-5.4 Mini
Benchmark	Claude Opus 4.8	GPT-5.4 Mini	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

Claude Opus 4.8$0.018Fits in one request

GPT-5.4 Mini$0.0030Fits in one request

GPT-5.4 Mini costs less

Repository review

50K input + 3K output tokens

Claude Opus 4.8$0.325Fits in one request

GPT-5.4 Mini$0.051Fits in one request

GPT-5.4 Mini costs less

Agent loop

200K input + 10K output tokens

Claude Opus 4.8$1.25Fits in one request

GPT-5.4 Mini$0.195Fits in one request

GPT-5.4 Mini 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 Claude Opus 4.8 and GPT-5.4 Mini
Specification	Anthropic Claude Opus 4.8	OpenAI GPT-5.4 Mini
Maker	Anthropic	OpenAI
Weights	No public weights listed	No public weights listed
Released	May 27, 2026	March 17, 2026
Context window	1M	400K
Input $/Mtok	$5.00	$0.75
Output $/Mtok	$25.00	$4.50
Coding index	65.0	35.8
Independent measurements behind it	6	5
Cost per solved task	$2.17	$0.70
Source	OpenRouter listing	OpenRouter listing

Questions

Is Claude Opus 4.8 better than GPT-5.4 Mini for coding?

Pick Claude Opus 4.8 for coding: it scores 65.0 on the Vibe Coding Index against 35.8 for GPT-5.4 Mini. Pick GPT-5.4 Mini for output tokens 5.6x cheaper ($4.50 vs $25.00 per million) and a lower cost per solved task ($0.70 against $2.17).

Which is cheaper, Claude Opus 4.8 or GPT-5.4 Mini?

Claude Opus 4.8 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-5.4 Mini costs $0.75 and $4.50. Per solved SWE-bench Verified task, Claude Opus 4.8 costs $2.17 and GPT-5.4 Mini $0.70.

Which has the bigger context window, Claude Opus 4.8 or GPT-5.4 Mini?

Claude Opus 4.8, with 1M tokens against 400K.

How many benchmarks is the Claude Opus 4.8 vs GPT-5.4 Mini comparison based on?

85 benchmarks have a result for both models as of October 21, 2026. Claude Opus 4.8's coding index rests on 6 independent benchmarks and GPT-5.4 Mini'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/claude-opus-4-8-vs-gpt-5-4-mini" title="Claude Opus 4.8 vs GPT-5.4 Mini, 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/claude-opus-4-8-vs-gpt-5-4-mini">LLMBenchmark.ai, Claude Opus 4.8 vs GPT-5.4 Mini</a></p>

Claude Opus 4.8 or GPT-5.4 Mini: which to pick

Which one for your work

Coding

Intelligence

Agentic work

Reasoning

Math

Knowledge

Web design

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