Claude Opus 5.5 vs GPT-5.4 Mini
Claude Opus 5.5 leads GPT-5.4 Mini on coding, 88.8 to 35.8 on the Vibe Coding Index, and GPT-5.4 Mini costs 4.4x less per output token ($4.50 vs $20.00 per million). Updated October 21, 2026; 63 benchmarks have a result for both.
Claude Opus 5.5 or GPT-5.4 Mini: which to pick
Pick Claude Opus 5.5 for coding: it scores 88.8 on the Vibe Coding Index against 35.8 for GPT-5.4 Mini.
Pick GPT-5.4 Mini for output tokens 4.4x cheaper ($4.50 vs $20.00 per million) and work measured by SAGE (Vals AI), where it leads.
- Shared results
- 63
- Claude Opus 5.5 only
- 46
- GPT-5.4 Mini only
- 26
- 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 5.5
Claude Opus 5.5 scores 88.8 against 35.8.
5 benchmarks behind each score
Intelligence
AA Intelligence Index
Claude Opus 5.5
Claude Opus 5.5 leads, 57.6 against 24.1.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only GPT-5.4 Mini has a AA Agentic Index result, 17.9.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5.4 Mini has a GPQA Diamond result, 83.1%.
Directional only
Math
AIME
No call
Only GPT-5.4 Mini has a AIME result, 95.6%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.4 Mini has a MMLU Pro result, 84.5%.
Directional only
Web design
Design Arena: websites
No call
Only Claude Opus 5.5 has a Design Arena: websites result, 1336.
Directional only
General
Vals Index
No call
Only Claude Opus 5.5 has a Vals Index result, 70.2%.
Directional only
Long documents
Listed context window
Claude Opus 5.5
Claude Opus 5.5 takes 1M tokens against 400K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GPT-5.4 Mini
Claude Opus 5.5 $0.014, GPT-5.4 Mini $0.0030 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.4 Mini has the number, $0.70.
Directional only
At a glance
- Coding index
- Claude Opus 5.5 88.8 from 5 benchmarks, GPT-5.4 Mini 35.8 from 5 benchmarks.
- Widest category
- Math, 41.1 points apart on average over 6 shared benchmarks.
- Widest single gap
- FrontierMath, tier 4: Claude Opus 5.5 95.0% against GPT-5.4 Mini 9.8%, 85.2 points apart on a 0 to 100 scale.
- Output price
- Claude Opus 5.5 $20.00 against GPT-5.4 Mini $4.50 per million output tokens, 4.4x apart.
- Context window
- Claude Opus 5.5 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.
| Category | Claude Opus 5.5 | Bars | GPT-5.4 Mini | Ties |
|---|---|---|---|---|
| Coding | 13 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 4 | 0 | ||
| Reasoning | 11 | 0 | ||
| Math | 6 | 0 | ||
| Knowledge | 5 | 0 | ||
| General | 21 | 1 | ||
| All shared boards | 62 | 1 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index130
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| LiveBench coding | 96.3% | Not measured | Only Claude Opus 5.5 measured |
| LMArena WebDev | 1807 | 1397 | Claude Opus 5.5 |
| AA Coding Index | Not measured | 56.1 | Only GPT-5.4 Mini measured |
| LiveCodeBench (Vals AI) | Not measured | 81.5% | Only GPT-5.4 Mini measured |
| SWE-bench Verified (Vals AI) | Not measured | 73.0% | Only GPT-5.4 Mini measured |
| Terminal-Bench 4 (Vals AI) | 65.2% | 2.5% | Claude Opus 5.5 |
| Vibe Code Bench (Vals AI) | 90.3% | 48.0% | Claude Opus 5.5 |
| Code migration (Vals AI) | 66.7% | 12.9% | Claude Opus 5.5 |
| ProgramBench (Vals AI) | 18.5% | 0.0% | Claude Opus 5.5 |
| IOI (Vals AI) | 95.1% | Not measured | Only Claude Opus 5.5 measured |
| Terminal-Bench 2.1 (Vals AI) | 87.6% | 54.7% | Claude Opus 5.5 |
| SWE-bench Multimodal (Vals AI) | 70.2% | Not measured | Only Claude Opus 5.5 measured |
| Vibe Code Bench 1-100 (Vals AI) | 30.4% | Not measured | Only Claude Opus 5.5 measured |
| CyberBench (Vals AI) | 74.6% | Not measured | Only Claude Opus 5.5 measured |
| SciCode | 66.9% | 52.1% | Claude Opus 5.5 |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 52.3% | Only GPT-5.4 Mini measured |
| FrontierCode | 54.6% | 27.0% | Claude Opus 5.5 |
| FrontierSWE | 62.3% | Not measured | Only Claude Opus 5.5 measured |
| WeirdML | Not measured | 60.3% | Only GPT-5.4 Mini measured |
| MirrorCode | 77.4% | Not measured | Only Claude Opus 5.5 measured |
| ALE-Bench | 2147.4 | 1188.6 | Claude Opus 5.5 |
| LMArena Text: coding | 1545 | 1438 | Claude Opus 5.5 |
| LMArena Text: software and IT | 1538 | 1436 | Claude Opus 5.5 |
| LMArena WebDev: React | 1855 | 1386 | Claude Opus 5.5 |
| LMArena WebDev: HTML | 1753 | 1419 | Claude Opus 5.5 |
| LMArena WebDev: from an image | 1753 | Not measured | Only Claude Opus 5.5 measured |
| SciCode | 18.5% | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 12 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| AA Intelligence Index | 57.6 | 24.1 | Claude Opus 5.5 |
| Epoch Capabilities Index | 168.8 | 148.7 | Claude Opus 5.5 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 17.9 | Only GPT-5.4 Mini measured |
| Design Arena: agents building web apps | 1317 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: agents building mobile apps | 1310 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: spreadsheets | 1352 | Not measured | Only Claude Opus 5.5 measured |
| LiveBench research | 85.0% | Not measured | Only Claude Opus 5.5 measured |
| LiveBench browser | 80.0% | Not measured | Only Claude Opus 5.5 measured |
| LiveBench long agentic | 85.0% | Not measured | Only Claude Opus 5.5 measured |
| SREBench (Vals AI) | 33.6% | Not measured | Only Claude Opus 5.5 measured |
| Finance Agent (Vals AI) | 58.6% | 45.4% | Claude Opus 5.5 |
| Tax Agent Bench (Vals AI) | 45.6% | 8.8% | Claude Opus 5.5 |
| Legal Research Bench (Vals AI) | 50.5% | 12.5% | Claude Opus 5.5 |
| Harvey's Legal Agent Benchmark (Vals AI) | 9.2% | Not measured | Only Claude Opus 5.5 measured |
| Excel Modeling Benchmark (Vals AI) | 75.9% | 45.4% | Claude Opus 5.5 |
| BioMysteryBench (Vals AI) | 79.3% | Not measured | Only Claude Opus 5.5 measured |
| Terminal-Bench-Science (Vals AI) | 47.1% | Not measured | Only Claude Opus 5.5 measured |
| Vals Science Index | 58.7% | Not measured | Only Claude Opus 5.5 measured |
| Time Horizon Index (Vals AI) | 91.3% | Not measured | Only Claude Opus 5.5 measured |
| CUA-bench (Vals AI) | 14.0% | Not measured | Only Claude Opus 5.5 measured |
| Robo-bench (Vals AI) | 14.6% | Not measured | Only Claude Opus 5.5 measured |
| τ²-Bench Telecom | Not measured | 83.3% | Only GPT-5.4 Mini measured |
| APEX-Agents | 73.5% | Not measured | Only Claude Opus 5.5 measured |
| Remote Labor Index | 21.3% | Not measured | Only Claude Opus 5.5 measured |
| DeepResearch Bench | Not measured | 36.3% | Only GPT-5.4 Mini measured |
| SpireClimb | 3.3% | Not measured | Only Claude Opus 5.5 measured |
| Vending-Bench 2 | $9,235 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Search | 1213 | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 15 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index110
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 83.1% | Only GPT-5.4 Mini measured |
| LiveBench reasoning | 92.2% | Not measured | Only Claude Opus 5.5 measured |
| MysteryMechanism (Vals AI) | 49.5% | Not measured | Only Claude Opus 5.5 measured |
| MMMU Pro (Vals AI) | Not measured | 79.3% | Only GPT-5.4 Mini measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 87.5% | Only GPT-5.4 Mini measured |
| Humanity's Last Exam (Artificial Analysis run) | 61.4% | 28.1% | Claude Opus 5.5 |
| GPQA Diamond (Epoch AI run) | 90.6% | 86.9% | Claude Opus 5.5 |
| ARC-AGI | 98.5% | 63.7% | Claude Opus 5.5 |
| ARC-AGI-2 | 93.3% | 18.9% | Claude Opus 5.5 |
| Chess puzzles | 67.0% | 23.0% | Claude Opus 5.5 |
| Mystery game puzzles | 71.0% | 11.0% | Claude Opus 5.5 |
| EBR-bench | 71.4% | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: hard prompts | 1536 | 1424 | Claude Opus 5.5 |
| LMArena Text: expert | 1549 | 1435 | Claude Opus 5.5 |
| LMArena Text: hard prompts in English | 1535 | 1420 | Claude Opus 5.5 |
| LMArena Vision: diagrams | 1345 | 1266 | Claude Opus 5.5 |
| LMArena Vision: homework | 1362 | 1285 | Claude Opus 5.5 |
| Kagi LLM Benchmark | Not measured | 37.9% | Only GPT-5.4 Mini measured |
| Neither model has a result on the other 4 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| AIME | Not measured | 95.6% | Only GPT-5.4 Mini measured |
| OTIS Mock AIME | 100.0% | 88.9% | Claude Opus 5.5 |
| FrontierMath, tiers 1 to 3 | 91.2% | 51.2% | Claude Opus 5.5 |
| FrontierMath, tier 4 | 95.0% | 9.8% | Claude Opus 5.5 |
| ProofBench | 100.0% | 21.0% | Claude Opus 5.5 |
| LMArena Text: math | 1518 | 1418 | Claude Opus 5.5 |
| LMArena Text: mathematical work | 1516 | 1429 | Claude Opus 5.5 |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 84.5% | Only GPT-5.4 Mini measured |
| CaseLaw (Vals AI) | Not measured | 51.7% | Only GPT-5.4 Mini measured |
| CorpFin (Vals AI) | Not measured | 60.9% | Only GPT-5.4 Mini measured |
| TaxEval (Vals AI) | Not measured | 71.2% | Only GPT-5.4 Mini measured |
| MortgageTax (Vals AI) | Not measured | 63.5% | Only GPT-5.4 Mini measured |
| MedCode (Vals AI) | 49.8% | Not measured | Only Claude Opus 5.5 measured |
| SimpleQA Verified | 72.2% | 29.4% | Claude Opus 5.5 |
| LMArena Text: science | 1524 | 1417 | Claude Opus 5.5 |
| LMArena Text: medicine and healthcare | 1509 | 1395 | Claude Opus 5.5 |
| LMArena Text: legal and government | 1518 | 1414 | Claude Opus 5.5 |
| LMArena Text: business and finance | 1494 | 1418 | Claude Opus 5.5 |
| FACTS Benchmark Suite | 65.3% | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 6 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| Design Arena: websites | 1336 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: UI components | 1392 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: game dev | 1458 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: data viz | 1397 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: 3D | 1477 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: all code categories | 1390 | Not measured | Only Claude Opus 5.5 measured |
| Design Arena: SVG | 1333 | Not measured | Only Claude Opus 5.5 measured |
| LiveBench design | 86.9% | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index211
| Benchmark | Claude Opus 5.5 | GPT-5.4 Mini | Leader |
|---|---|---|---|
| Vals Index | 70.2% | Not measured | Only Claude Opus 5.5 measured |
| LiveBench data analysis | 80.3% | Not measured | Only Claude Opus 5.5 measured |
| Public Benefits Bench (Vals AI) | 70.6% | Not measured | Only Claude Opus 5.5 measured |
| SAGE (Vals AI) | 45.8% | 50.8% | GPT-5.4 Mini |
| MedScribe (Vals AI) | 91.4% | Not measured | Only Claude Opus 5.5 measured |
| Vals Multimodal Index | Not measured | 54.2% | Only GPT-5.4 Mini measured |
| IFBench | Not measured | 73.3% | Only GPT-5.4 Mini measured |
| Long context reasoning (AA-LCR) | 84.7% | 77.0% | Claude Opus 5.5 |
| DTBench | 98.9% | 80.0% | Claude Opus 5.5 |
| LMCA | 68.2% | 40.8% | Claude Opus 5.5 |
| Furniture assembly | 83.3% | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text | 1516 | 1412 | Claude Opus 5.5 |
| LMArena Text: instruction following | 1537 | 1405 | Claude Opus 5.5 |
| LMArena Text: creative writing | 1535 | 1370 | Claude Opus 5.5 |
| LMArena Vision | 1321 | 1245 | Claude Opus 5.5 |
| LMArena Text: English | 1519 | 1412 | Claude Opus 5.5 |
| LMArena Text: non-English | 1507 | 1405 | Claude Opus 5.5 |
| LMArena Text: Chinese | 1582 | 1446 | Claude Opus 5.5 |
| LMArena Text: German | Not measured | 1408 | Only GPT-5.4 Mini measured |
| LMArena Text: Japanese | 1534 | 1375 | Claude Opus 5.5 |
| LMArena Text: Korean | Not measured | 1370 | Only GPT-5.4 Mini measured |
| LMArena Text: French | 1493 | 1441 | Claude Opus 5.5 |
| LMArena Text: Spanish | 1508 | 1405 | Claude Opus 5.5 |
| LMArena Text: Russian | 1517 | 1417 | Claude Opus 5.5 |
| LMArena Text: Polish | Not measured | 1415 | Only GPT-5.4 Mini measured |
| LMArena Text: multi-turn | 1501 | 1429 | Claude Opus 5.5 |
| LMArena Text: longer queries | 1532 | 1408 | Claude Opus 5.5 |
| LMArena Text: writing and language | 1532 | 1393 | Claude Opus 5.5 |
| LMArena Text: entertainment and media | 1515 | 1373 | Claude Opus 5.5 |
| LMArena Vision: English | 1317 | 1244 | Claude Opus 5.5 |
| LMArena Vision: Chinese | Not measured | 1286 | Only GPT-5.4 Mini measured |
| LMArena Vision: reading text | 1326 | 1253 | Claude Opus 5.5 |
| LMArena Vision: humor | Not measured | 1235 | Only GPT-5.4 Mini measured |
| LMArena Vision: recognizing things | Not measured | 1218 | Only GPT-5.4 Mini measured |
| LMArena Vision: creative writing | 1320 | 1222 | Claude Opus 5.5 |
| Neither model has a result on the other 11 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Claude Opus 5.5 | 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
GPT-5.4 Mini costs less
Repository review
50K input + 3K output tokens
GPT-5.4 Mini costs less
Agent loop
200K input + 10K output tokens
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.
| Specification | Anthropic Claude Opus 5.5 | OpenAI GPT-5.4 Mini |
|---|---|---|
| Maker | Anthropic | OpenAI |
| Weights | No public weights listed | No public weights listed |
| Released | September 22, 2026 | March 17, 2026 |
| Context window | 1M | 400K |
| Input $/Mtok | $4.00 | $0.75 |
| Output $/Mtok | $20.00 | $4.50 |
| Coding index | 88.8 | 35.8 |
| Independent measurements behind it | 5 | 5 |
| Cost per solved task | Not computable | $0.70 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Opus 5.5 better than GPT-5.4 Mini for coding?
Pick Claude Opus 5.5 for coding: it scores 88.8 on the Vibe Coding Index against 35.8 for GPT-5.4 Mini. Pick GPT-5.4 Mini for output tokens 4.4x cheaper ($4.50 vs $20.00 per million) and work measured by SAGE (Vals AI), where it leads.
Which is cheaper, Claude Opus 5.5 or GPT-5.4 Mini?
Claude Opus 5.5 costs $4.00 per million input tokens and $20.00 per million output tokens; GPT-5.4 Mini costs $0.75 and $4.50. Cost per solved task is known for GPT-5.4 Mini only.
Which has the bigger context window, Claude Opus 5.5 or GPT-5.4 Mini?
Claude Opus 5.5, with 1M tokens against 400K.
How many benchmarks is the Claude Opus 5.5 vs GPT-5.4 Mini comparison based on?
63 benchmarks have a result for both models as of October 21, 2026. Claude Opus 5.5's coding index rests on 5 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.