GPT-5.4 Mini vs GPT-5.6 Sol
GPT-5.6 Sol leads GPT-5.4 Mini on coding, 73.5 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; 85 benchmarks have a result for both.
GPT-5.4 Mini or GPT-5.6 Sol: which to pick
Pick GPT-5.6 Sol for coding: it scores 73.5 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 a lower cost per solved task ($0.70 against $1.20).
- Shared results
- 85
- GPT-5.4 Mini only
- 4
- GPT-5.6 Sol only
- 54
- 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
GPT-5.6 Sol
GPT-5.6 Sol scores 73.5 against 35.8.
5 and 6 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-5.6 Sol
GPT-5.6 Sol leads, 47.0 against 24.1.
Same board, both measured
Agentic work
AA Agentic Index
GPT-5.6 Sol
GPT-5.6 Sol leads, 50.2 against 17.9.
Same board, both measured
Reasoning
GPQA Diamond
GPT-5.6 Sol
GPT-5.6 Sol leads, 95.2% 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
GPT-5.6 Sol
GPT-5.6 Sol leads, 89.1% against 84.5%.
Same board, both measured
Web design
Design Arena: websites
No call
Only GPT-5.6 Sol has a Design Arena: websites result, 1321.
Directional only
Long documents
Listed context window
GPT-5.6 Sol
GPT-5.6 Sol takes 1.05M tokens against 400K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GPT-5.4 Mini
GPT-5.4 Mini $0.0030, GPT-5.6 Sol $0.014 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
GPT-5.4 Mini
GPT-5.4 Mini $0.70, GPT-5.6 Sol $1.20.
Published attempt cost and pass rate
At a glance
- Coding index
- GPT-5.4 Mini 35.8 from 5 benchmarks, GPT-5.6 Sol 73.5 from 6 benchmarks.
- Widest category
- Math, 34.1 points apart on average over 6 shared benchmarks.
- Widest single gap
- ARC-AGI-2: GPT-5.4 Mini 18.9% against GPT-5.6 Sol 92.5%, 73.6 points apart on a 0 to 100 scale.
- Output price
- GPT-5.4 Mini $4.50 against GPT-5.6 Sol $20.00 per million output tokens, 4.4x apart.
- Context window
- GPT-5.4 Mini takes 400K tokens, GPT-5.6 Sol 1.05M.
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 | GPT-5.4 Mini | Bars | GPT-5.6 Sol | Ties |
|---|---|---|---|---|
| Coding | 0 | 18 | ||
| Intelligence | 0 | 2 | ||
| Agentic work | 0 | 6 | ||
| Reasoning | 0 | 15 | ||
| Math | 0 | 6 | ||
| Knowledge | 0 | 9 | ||
| General | 1 | 28 | ||
| All shared boards | 1 | 84 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index018
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 89.7% | Only GPT-5.6 Sol measured |
| LMArena WebDev | 1397 | 1616 | GPT-5.6 Sol |
| AA Coding Index | 56.1 | 77.4 | GPT-5.6 Sol |
| LiveCodeBench (Vals AI) | 81.5% | 82.6% | GPT-5.6 Sol |
| SWE-bench Verified (Vals AI) | 73.0% | 96.2% | GPT-5.6 Sol |
| Terminal-Bench 4 (Vals AI) | 2.5% | 37.9% | GPT-5.6 Sol |
| Vibe Code Bench (Vals AI) | 48.0% | 80.5% | GPT-5.6 Sol |
| Code migration (Vals AI) | 12.9% | 52.9% | GPT-5.6 Sol |
| ProgramBench (Vals AI) | 0.0% | 1.5% | GPT-5.6 Sol |
| IOI (Vals AI) | Not measured | 91.2% | Only GPT-5.6 Sol measured |
| Terminal-Bench 2.1 (Vals AI) | 54.7% | 85.8% | GPT-5.6 Sol |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 20.0% | Only GPT-5.6 Sol measured |
| SkillsBench (Vals AI) | Not measured | 54.1% | Only GPT-5.6 Sol measured |
| CyberBench (Vals AI) | Not measured | 76.3% | Only GPT-5.6 Sol measured |
| SciCode | 52.1% | 57.8% | GPT-5.6 Sol |
| Terminal-Bench Hard (Artificial Analysis run) | 52.3% | 65.9% | GPT-5.6 Sol |
| FrontierCode | 27.0% | 47.5% | GPT-5.6 Sol |
| FrontierSWE | Not measured | 32.2% | Only GPT-5.6 Sol measured |
| GSO-Bench | Not measured | 76.5% | Only GPT-5.6 Sol measured |
| WeirdML | 60.3% | 88.8% | GPT-5.6 Sol |
| MirrorCode | Not measured | 20.0% | Only GPT-5.6 Sol measured |
| ALE-Bench | 1188.6 | 2176.9 | GPT-5.6 Sol |
| LMArena Text: coding | 1438 | 1496 | GPT-5.6 Sol |
| LMArena Text: software and IT | 1436 | 1489 | GPT-5.6 Sol |
| LMArena WebDev: React | 1386 | 1620 | GPT-5.6 Sol |
| LMArena WebDev: HTML | 1419 | 1618 | GPT-5.6 Sol |
| LMArena WebDev: from an image | Not measured | 1609 | Only GPT-5.6 Sol measured |
| SciCode | Not measured | 10.8% | Only GPT-5.6 Sol measured |
| Neither model has a result on the other 11 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| AA Intelligence Index | 24.1 | 47.0 | GPT-5.6 Sol |
| Epoch Capabilities Index | 148.7 | 162.8 | GPT-5.6 Sol |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| AA Agentic Index | 17.9 | 50.2 | GPT-5.6 Sol |
| Design Arena: agents building web apps | Not measured | 1245 | Only GPT-5.6 Sol measured |
| Design Arena: agents building full-stack apps | Not measured | 1261 | Only GPT-5.6 Sol measured |
| Design Arena: agents building mobile apps | Not measured | 1248 | Only GPT-5.6 Sol measured |
| Design Arena: agents building Android apps | Not measured | 1239 | Only GPT-5.6 Sol measured |
| Design Arena: agents building Godot games | Not measured | 1260 | Only GPT-5.6 Sol measured |
| Design Arena: agents building games | Not measured | 1267 | Only GPT-5.6 Sol measured |
| Design Arena: HTML slides | Not measured | 1286 | Only GPT-5.6 Sol measured |
| Design Arena: PowerPoint slides | Not measured | 1241 | Only GPT-5.6 Sol measured |
| LiveBench research | Not measured | 54.5% | Only GPT-5.6 Sol measured |
| LiveBench browser | Not measured | 88.0% | Only GPT-5.6 Sol measured |
| LiveBench long agentic | Not measured | 48.5% | Only GPT-5.6 Sol measured |
| SREBench (Vals AI) | Not measured | 30.5% | Only GPT-5.6 Sol measured |
| Finance Agent (Vals AI) | 45.4% | 53.8% | GPT-5.6 Sol |
| Tax Agent Bench (Vals AI) | 8.8% | 31.1% | GPT-5.6 Sol |
| Legal Research Bench (Vals AI) | 12.5% | 48.1% | GPT-5.6 Sol |
| Excel Modeling Benchmark (Vals AI) | 45.4% | 72.3% | GPT-5.6 Sol |
| BioMysteryBench (Vals AI) | Not measured | 71.1% | Only GPT-5.6 Sol measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 20.0% | Only GPT-5.6 Sol measured |
| Vals Science Index | Not measured | 41.5% | Only GPT-5.6 Sol measured |
| Time Horizon Index (Vals AI) | Not measured | 23.8% | Only GPT-5.6 Sol measured |
| CUA-bench (Vals AI) | Not measured | 8.3% | Only GPT-5.6 Sol measured |
| τ²-Bench Telecom | 83.3% | 85.1% | GPT-5.6 Sol |
| OSWorld 2.0 | Not measured | 27.3% | Only GPT-5.6 Sol measured |
| BALROG | Not measured | 60.0% | Only GPT-5.6 Sol measured |
| DeepResearch Bench | 36.3% | Not measured | Only GPT-5.4 Mini measured |
| PostTrainBench | Not measured | 36.2% | Only GPT-5.6 Sol measured |
| Vending-Bench 2 | Not measured | $9,619 | Only GPT-5.6 Sol measured |
| LMArena Search | Not measured | 1261 | Only GPT-5.6 Sol measured |
| Neither model has a result on the other 12 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index015
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| GPQA Diamond | 83.1% | 95.2% | GPT-5.6 Sol |
| LiveBench reasoning | Not measured | 91.7% | Only GPT-5.6 Sol measured |
| MysteryMechanism (Vals AI) | Not measured | 33.3% | Only GPT-5.6 Sol measured |
| MMMU Pro (Vals AI) | 79.3% | 88.8% | GPT-5.6 Sol |
| GPQA Diamond (Artificial Analysis run) | 87.5% | 94.1% | GPT-5.6 Sol |
| Humanity's Last Exam (Artificial Analysis run) | 28.1% | 49.5% | GPT-5.6 Sol |
| GPQA Diamond (Epoch AI run) | 86.9% | 93.5% | GPT-5.6 Sol |
| ARC-AGI | 63.7% | 97.5% | GPT-5.6 Sol |
| ARC-AGI-2 | 18.9% | 92.5% | GPT-5.6 Sol |
| SimpleBench | Not measured | 64.8% | Only GPT-5.6 Sol measured |
| Chess puzzles | 23.0% | 59.0% | GPT-5.6 Sol |
| Mystery game puzzles | 11.0% | 58.0% | GPT-5.6 Sol |
| EBR-bench | Not measured | 44.8% | Only GPT-5.6 Sol measured |
| LMArena Text: hard prompts | 1424 | 1483 | GPT-5.6 Sol |
| LMArena Text: expert | 1435 | 1517 | GPT-5.6 Sol |
| LMArena Text: hard prompts in English | 1420 | 1477 | GPT-5.6 Sol |
| LMArena Vision: diagrams | 1266 | 1305 | GPT-5.6 Sol |
| LMArena Vision: homework | 1285 | 1305 | GPT-5.6 Sol |
| Kagi LLM Benchmark | 37.9% | 67.0% | GPT-5.6 Sol |
| Neither model has a result on the other 3 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| AIME | 95.6% | Not measured | Only GPT-5.4 Mini measured |
| OTIS Mock AIME | 88.9% | 100.0% | GPT-5.6 Sol |
| FrontierMath, tiers 1 to 3 | 51.2% | 89.1% | GPT-5.6 Sol |
| FrontierMath, tier 4 | 9.8% | 82.9% | GPT-5.6 Sol |
| ProofBench | 21.0% | 83.0% | GPT-5.6 Sol |
| LMArena Text: math | 1418 | 1473 | GPT-5.6 Sol |
| LMArena Text: mathematical work | 1429 | 1496 | GPT-5.6 Sol |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index09
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| MMLU Pro | 84.5% | 89.1% | GPT-5.6 Sol |
| LegalBench (Vals AI) | Not measured | 87.0% | Only GPT-5.6 Sol measured |
| CaseLaw (Vals AI) | 51.7% | Not measured | Only GPT-5.4 Mini measured |
| CorpFin (Vals AI) | 60.9% | 64.4% | GPT-5.6 Sol |
| TaxEval (Vals AI) | 71.2% | 74.8% | GPT-5.6 Sol |
| MortgageTax (Vals AI) | 63.5% | 67.3% | GPT-5.6 Sol |
| MedCode (Vals AI) | Not measured | 44.0% | Only GPT-5.6 Sol measured |
| SimpleQA Verified | 29.4% | 69.7% | GPT-5.6 Sol |
| LMArena Text: science | 1417 | 1455 | GPT-5.6 Sol |
| LMArena Text: medicine and healthcare | 1395 | 1431 | GPT-5.6 Sol |
| LMArena Text: legal and government | 1414 | 1476 | GPT-5.6 Sol |
| LMArena Text: business and finance | 1418 | 1457 | GPT-5.6 Sol |
| FACTS Benchmark Suite | Not measured | 52.4% | Only GPT-5.6 Sol 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
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1321 | Only GPT-5.6 Sol measured |
| Design Arena: UI components | Not measured | 1336 | Only GPT-5.6 Sol measured |
| Design Arena: game dev | Not measured | 1368 | Only GPT-5.6 Sol measured |
| Design Arena: data viz | Not measured | 1322 | Only GPT-5.6 Sol measured |
| Design Arena: 3D | Not measured | 1386 | Only GPT-5.6 Sol measured |
| Design Arena: all code categories | Not measured | 1341 | Only GPT-5.6 Sol measured |
| Design Arena: SVG | Not measured | 1316 | Only GPT-5.6 Sol measured |
| Design Arena: ASCII art | Not measured | 1304 | Only GPT-5.6 Sol measured |
| LiveBench design | Not measured | 81.6% | Only GPT-5.6 Sol measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index128
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | Leader |
|---|---|---|---|
| LiveBench data analysis | Not measured | 79.8% | Only GPT-5.6 Sol measured |
| Public Benefits Bench (Vals AI) | Not measured | 66.5% | Only GPT-5.6 Sol measured |
| SAGE (Vals AI) | 50.8% | 52.6% | GPT-5.6 Sol |
| MedScribe (Vals AI) | Not measured | 85.2% | Only GPT-5.6 Sol measured |
| Vals Multimodal Index | 54.2% | 72.6% | GPT-5.6 Sol |
| IFBench | 73.3% | 72.7% | GPT-5.4 Mini |
| Long context reasoning (AA-LCR) | 77.0% | 84.0% | GPT-5.6 Sol |
| DTBench | 80.0% | 95.7% | GPT-5.6 Sol |
| LMCA | 40.8% | 58.5% | GPT-5.6 Sol |
| Surface Evolver Bench | Not measured | 93.1% | Only GPT-5.6 Sol measured |
| Furniture assembly | Not measured | 56.7% | Only GPT-5.6 Sol measured |
| LMArena Text | 1412 | 1457 | GPT-5.6 Sol |
| LMArena Text: instruction following | 1405 | 1483 | GPT-5.6 Sol |
| LMArena Text: creative writing | 1370 | 1449 | GPT-5.6 Sol |
| LMArena Vision | 1245 | 1281 | GPT-5.6 Sol |
| LMArena Text: English | 1412 | 1457 | GPT-5.6 Sol |
| LMArena Text: non-English | 1405 | 1451 | GPT-5.6 Sol |
| LMArena Text: Chinese | 1446 | 1525 | GPT-5.6 Sol |
| LMArena Text: German | 1408 | 1476 | GPT-5.6 Sol |
| LMArena Text: Japanese | 1375 | 1472 | GPT-5.6 Sol |
| LMArena Text: Korean | 1370 | 1442 | GPT-5.6 Sol |
| LMArena Text: French | 1441 | 1472 | GPT-5.6 Sol |
| LMArena Text: Spanish | 1405 | 1443 | GPT-5.6 Sol |
| LMArena Text: Russian | 1417 | 1471 | GPT-5.6 Sol |
| LMArena Text: Polish | 1415 | 1459 | GPT-5.6 Sol |
| LMArena Text: multi-turn | 1429 | 1461 | GPT-5.6 Sol |
| LMArena Text: longer queries | 1408 | 1481 | GPT-5.6 Sol |
| LMArena Text: writing and language | 1393 | 1462 | GPT-5.6 Sol |
| LMArena Text: entertainment and media | 1373 | 1450 | GPT-5.6 Sol |
| LMArena Vision: English | 1244 | 1280 | GPT-5.6 Sol |
| LMArena Vision: Chinese | 1286 | 1313 | GPT-5.6 Sol |
| LMArena Vision: reading text | 1253 | 1290 | GPT-5.6 Sol |
| LMArena Vision: humor | 1235 | 1246 | GPT-5.6 Sol |
| LMArena Vision: recognizing things | 1218 | Not measured | Only GPT-5.4 Mini measured |
| LMArena Vision: creative writing | 1222 | 1283 | GPT-5.6 Sol |
| LMArena Document | Not measured | 1483 | Only GPT-5.6 Sol 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
| Benchmark | GPT-5.4 Mini | GPT-5.6 Sol | 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 | OpenAI GPT-5.4 Mini | OpenAI GPT-5.6 Sol |
|---|---|---|
| Maker | OpenAI | OpenAI |
| Weights | No public weights listed | No public weights listed |
| Released | March 17, 2026 | July 9, 2026 |
| Context window | 400K | 1.05M |
| Input $/Mtok | $0.75 | $4.00 |
| Output $/Mtok | $4.50 | $20.00 |
| Coding index | 35.8 | 73.5 |
| Independent measurements behind it | 5 | 6 |
| Cost per solved task | $0.70 | $1.20 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-5.4 Mini better than GPT-5.6 Sol for coding?
Pick GPT-5.6 Sol for coding: it scores 73.5 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 a lower cost per solved task ($0.70 against $1.20).
Which is cheaper, GPT-5.4 Mini or GPT-5.6 Sol?
GPT-5.4 Mini costs $0.75 per million input tokens and $4.50 per million output tokens; GPT-5.6 Sol costs $4.00 and $20.00. Per solved SWE-bench Verified task, GPT-5.4 Mini costs $0.70 and GPT-5.6 Sol $1.20.
Which has the bigger context window, GPT-5.4 Mini or GPT-5.6 Sol?
GPT-5.6 Sol, with 1.05M tokens against 400K.
How many benchmarks is the GPT-5.4 Mini vs GPT-5.6 Sol comparison based on?
85 benchmarks have a result for both models as of October 21, 2026. GPT-5.4 Mini's coding index rests on 5 independent benchmarks and GPT-5.6 Sol's on 6. Where only one model has a result, the page says so instead of naming a winner.