GPT-6 Sol vs Mercury 3
Mercury 3 costs 28.6x less per output token than GPT-6 Sol ($0.35 vs $10.00 per million). Updated October 21, 2026; 5 benchmarks have a result for both.
vs boards won of 5 shared
GPT-6 Sol or Mercury 3: which to pick
Only GPT-6 Sol has a coding score, 76.0 on the Vibe Coding Index; Mercury 3 has none, so the two cannot be ranked on coding.
Pick Mercury 3 for output tokens 28.6x cheaper ($0.35 vs $10.00 per million).
Model AGPT-6 SolOpenAI76.0/1005 independent benchmarksNo cost per solved taskModel BMercury 3InceptionNot benchmarkedNo coding benchmarkNo cost per solved task
- Shared results
- 5
- GPT-6 Sol only
- 96
- Mercury 3 only
- 0
- Like-for-like categories
- 1 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
No call
Only GPT-6 Sol has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
No call
Only GPT-6 Sol has a AA Intelligence Index result, 47.6.
Directional only
Web design
Design Arena: websites
GPT-6 Sol
GPT-6 Sol leads, 1292 against 1244.
Same board, both measured
Long documents
Listed context window
GPT-6 Sol
GPT-6 Sol takes 1.05M tokens against 1M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Mercury 3
GPT-6 Sol $0.0070, Mercury 3 $0.0003 per turn.
List prices
At a glance
- Coding index
- GPT-6 Sol 76.0 from 5 benchmarks, Mercury 3 not benchmarked.
- Widest category
- Web design, 7.0 points apart on average over 5 shared benchmarks.
- Widest single gap
- Design Arena: game dev: GPT-6 Sol 1314 against Mercury 3 1227, 10.9 points apart on a 0 to 100 scale.
- Output price
- GPT-6 Sol $10.00 against Mercury 3 $0.35 per million output tokens, 28.6x apart.
- Context window
- GPT-6 Sol takes 1.05M tokens, Mercury 3 1M.
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-6 Sol | Bars | Mercury 3 | Ties |
|---|---|---|---|---|
| Web design | 5 | 0 | ||
| All shared boards | 5 | 0 |
CodingDirectional onlySix of these boards feed the Vibe Coding Index17 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| LiveBench coding | 93.1% | Not measured | Only GPT-6 Sol measured |
| LMArena WebDev | 1683 | Not measured | Only GPT-6 Sol measured |
| Terminal-Bench 4 (Vals AI) | 44.4% | Not measured | Only GPT-6 Sol measured |
| Vibe Code Bench (Vals AI) | 87.8% | Not measured | Only GPT-6 Sol measured |
| Code migration (Vals AI) | 57.2% | Not measured | Only GPT-6 Sol measured |
| ProgramBench (Vals AI) | 2.0% | Not measured | Only GPT-6 Sol measured |
| IOI (Vals AI) | 82.6% | Not measured | Only GPT-6 Sol measured |
| Terminal-Bench 2.1 (Vals AI) | 83.2% | Not measured | Only GPT-6 Sol measured |
| CyberBench (Vals AI) | 78.0% | Not measured | Only GPT-6 Sol measured |
| SciCode | 57.6% | Not measured | Only GPT-6 Sol measured |
| FrontierCode | 49.3% | Not measured | Only GPT-6 Sol measured |
| ALE-Bench | 2462.0 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: coding | 1448 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: software and IT | 1433 | Not measured | Only GPT-6 Sol measured |
| LMArena WebDev: React | 1685 | Not measured | Only GPT-6 Sol measured |
| LMArena WebDev: HTML | 1685 | Not measured | Only GPT-6 Sol measured |
| LMArena WebDev: from an image | 1660 | Not measured | Only GPT-6 Sol measured |
| Neither model has a result on the other 22 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| AA Intelligence Index | 47.6 | Not measured | Only GPT-6 Sol measured |
| Epoch Capabilities Index | 163.7 | Not measured | Only GPT-6 Sol measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index17 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| Design Arena: agents building web apps | 1231 | Not measured | Only GPT-6 Sol measured |
| Design Arena: agents building full-stack apps | 1238 | Not measured | Only GPT-6 Sol measured |
| Design Arena: agents building mobile apps | 1214 | Not measured | Only GPT-6 Sol measured |
| LiveBench research | 69.3% | Not measured | Only GPT-6 Sol measured |
| LiveBench browser | 92.0% | Not measured | Only GPT-6 Sol measured |
| LiveBench long agentic | 66.9% | Not measured | Only GPT-6 Sol measured |
| SREBench (Vals AI) | 30.5% | Not measured | Only GPT-6 Sol measured |
| Finance Agent (Vals AI) | 49.0% | Not measured | Only GPT-6 Sol measured |
| Tax Agent Bench (Vals AI) | 15.0% | Not measured | Only GPT-6 Sol measured |
| Legal Research Bench (Vals AI) | 28.9% | Not measured | Only GPT-6 Sol measured |
| Excel Modeling Benchmark (Vals AI) | 71.5% | Not measured | Only GPT-6 Sol measured |
| BioMysteryBench (Vals AI) | 74.8% | Not measured | Only GPT-6 Sol measured |
| Terminal-Bench-Science (Vals AI) | 30.0% | Not measured | Only GPT-6 Sol measured |
| Vals Science Index | 45.0% | Not measured | Only GPT-6 Sol measured |
| APEX-Agents | 54.3% | Not measured | Only GPT-6 Sol measured |
| SpireClimb | 7.5% | Not measured | Only GPT-6 Sol measured |
| Vending-Bench 2 | $14,428 | Not measured | Only GPT-6 Sol measured |
| Neither model has a result on the other 24 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index14 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| LiveBench reasoning | 88.7% | Not measured | Only GPT-6 Sol measured |
| MysteryMechanism (Vals AI) | 30.2% | Not measured | Only GPT-6 Sol measured |
| Humanity's Last Exam (Artificial Analysis run) | 47.9% | Not measured | Only GPT-6 Sol measured |
| GPQA Diamond (Epoch AI run) | 94.3% | Not measured | Only GPT-6 Sol measured |
| ARC-AGI | 95.5% | Not measured | Only GPT-6 Sol measured |
| ARC-AGI-2 | 89.6% | Not measured | Only GPT-6 Sol measured |
| Chess puzzles | 56.0% | Not measured | Only GPT-6 Sol measured |
| Mystery game puzzles | 56.0% | Not measured | Only GPT-6 Sol measured |
| EBR-bench | 53.3% | Not measured | Only GPT-6 Sol measured |
| LMArena Text: hard prompts | 1422 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: expert | 1437 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: hard prompts in English | 1420 | Not measured | Only GPT-6 Sol measured |
| LMArena Vision: diagrams | 1270 | Not measured | Only GPT-6 Sol measured |
| LMArena Vision: homework | 1286 | Not measured | Only GPT-6 Sol measured |
| Neither model has a result on the other 8 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| OTIS Mock AIME | 100.0% | Not measured | Only GPT-6 Sol measured |
| FrontierMath, tiers 1 to 3 | 89.8% | Not measured | Only GPT-6 Sol measured |
| FrontierMath, tier 4 | 90.0% | Not measured | Only GPT-6 Sol measured |
| ProofBench | 83.0% | Not measured | Only GPT-6 Sol measured |
| LMArena Text: math | 1407 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: mathematical work | 1424 | Not measured | Only GPT-6 Sol measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| MedCode (Vals AI) | 47.1% | Not measured | Only GPT-6 Sol measured |
| SimpleQA Verified | 60.7% | Not measured | Only GPT-6 Sol measured |
| LMArena Text: science | 1389 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: medicine and healthcare | 1366 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: legal and government | 1396 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: business and finance | 1380 | Not measured | Only GPT-6 Sol measured |
| FACTS Benchmark Suite | 72.5% | Not measured | Only GPT-6 Sol measured |
| Neither model has a result on the other 11 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| Design Arena: websites | 1292 | 1244 | GPT-6 Sol |
| Design Arena: UI components | 1292 | 1270 | GPT-6 Sol |
| Design Arena: game dev | 1314 | 1227 | GPT-6 Sol |
| Design Arena: data viz | 1269 | 1207 | GPT-6 Sol |
| Design Arena: 3D | 1318 | Not measured | Only GPT-6 Sol measured |
| Design Arena: all code categories | 1301 | 1240 | GPT-6 Sol |
| Design Arena: SVG | 1316 | Not measured | Only GPT-6 Sol measured |
| Design Arena: ASCII art | 1225 | Not measured | Only GPT-6 Sol measured |
| LiveBench design | 82.2% | Not measured | Only GPT-6 Sol measured |
GeneralDirectional onlyPublic boards, shown as published, kept out of the index29 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | Leader |
|---|---|---|---|
| LiveBench data analysis | 81.2% | Not measured | Only GPT-6 Sol measured |
| Public Benefits Bench (Vals AI) | 56.6% | Not measured | Only GPT-6 Sol measured |
| SAGE (Vals AI) | 44.8% | Not measured | Only GPT-6 Sol measured |
| MedScribe (Vals AI) | 82.0% | Not measured | Only GPT-6 Sol measured |
| Long context reasoning (AA-LCR) | 83.7% | Not measured | Only GPT-6 Sol measured |
| DTBench | 97.3% | Not measured | Only GPT-6 Sol measured |
| LMCA | 59.1% | Not measured | Only GPT-6 Sol measured |
| Furniture assembly | 58.3% | Not measured | Only GPT-6 Sol measured |
| LMArena Text | 1398 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: instruction following | 1413 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: creative writing | 1378 | Not measured | Only GPT-6 Sol measured |
| LMArena Vision | 1240 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: English | 1396 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: non-English | 1388 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: Chinese | 1412 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: German | 1379 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: Japanese | 1377 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: Korean | 1336 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: French | 1399 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: Spanish | 1387 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: Russian | 1402 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: Polish | 1372 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: multi-turn | 1414 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: longer queries | 1416 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: writing and language | 1389 | Not measured | Only GPT-6 Sol measured |
| LMArena Text: entertainment and media | 1373 | Not measured | Only GPT-6 Sol measured |
| LMArena Vision: English | 1229 | Not measured | Only GPT-6 Sol measured |
| LMArena Vision: reading text | 1258 | Not measured | Only GPT-6 Sol measured |
| LMArena Vision: creative writing | 1219 | Not measured | Only GPT-6 Sol measured |
| Neither model has a result on the other 17 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GPT-6 Sol | Mercury 3 | 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-6 Sol$0.0070Fits in one request
Mercury 3$0.0003Fits in one request
Mercury 3 costs less
Repository review
50K input + 3K output tokens
GPT-6 Sol$0.130Fits in one request
Mercury 3$0.0060Fits in one request
Mercury 3 costs less
Agent loop
200K input + 10K output tokens
GPT-6 Sol$0.500Fits in one request
Mercury 3$0.024Fits in one request
Mercury 3 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-6 Sol | Inception Mercury 3 |
|---|---|---|
| Maker | OpenAI | Inception |
| Weights | No public weights listed | No public weights listed |
| Released | September 22, 2026 | October 16, 2026 |
| Context window | 1.05M | 1M |
| Input $/Mtok | $2.00 | $0.10 |
| Output $/Mtok | $10.00 | $0.35 |
| Coding index | 76.0 | Not benchmarked |
| Independent measurements behind it | 5 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-6 Sol better than Mercury 3 for coding?
Only GPT-6 Sol has a coding score, 76.0 on the Vibe Coding Index; Mercury 3 has none, so the two cannot be ranked on coding. Pick Mercury 3 for output tokens 28.6x cheaper ($0.35 vs $10.00 per million).
Which is cheaper, GPT-6 Sol or Mercury 3?
GPT-6 Sol costs $2.00 per million input tokens and $10.00 per million output tokens; Mercury 3 costs $0.10 and $0.35. Cost per solved task is known for neither of them.
Which has the bigger context window, GPT-6 Sol or Mercury 3?
GPT-6 Sol, with 1.05M tokens against 1M.
How many benchmarks is the GPT-6 Sol vs Mercury 3 comparison based on?
5 benchmarks have a result for both models as of October 21, 2026. GPT-6 Sol's coding index rests on 5 independent benchmarks and Mercury 3's on 0. Where only one model has a result, the page says so instead of naming a winner.