GPT-6 Astra vs Mercury 2.5
GPT-6 Astra leads Mercury 2.5 on coding, 85.1 to 1.6 on the Vibe Coding Index, and Mercury 2.5 costs 333.3x less per output token ($0.15 vs $50.00 per million). Updated October 22, 2026; 19 benchmarks have a result for both.
vs boards won of 19 shared
GPT-6 Astra or Mercury 2.5: which to pick
Pick GPT-6 Astra for coding: it scores 85.1 on the Vibe Coding Index against 1.6 for Mercury 2.5.
Pick Mercury 2.5 for output tokens 333.3x cheaper ($0.15 vs $50.00 per million) and work measured by CyberBench (Vals AI), where it leads.
Model AGPT-6 AstraOpenAI85.1/1005 independent benchmarksNo cost per solved taskModel BMercury 2.5Inception1.6/1002 independent benchmarksNo cost per solved task
- Shared results
- 19
- GPT-6 Astra only
- 110
- Mercury 2.5 only
- 3
- 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-6 Astra
GPT-6 Astra scores 85.1 against 1.6.
5 and 2 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-6 Astra
GPT-6 Astra leads, 52.7 against 12.3.
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
Web design
Design Arena: websites
No call
Only GPT-6 Astra has a Design Arena: websites result, 1336.
Directional only
General
Vals Index
No call
Only GPT-6 Astra has a Vals Index result, 65.0%.
Directional only
Long documents
Listed context window
GPT-6 Astra
GPT-6 Astra takes 1.05M tokens against 260K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Mercury 2.5
GPT-6 Astra $0.035, Mercury 2.5 $0.0001 per turn.
List prices
At a glance
- Coding index
- GPT-6 Astra 85.1 from 5 benchmarks, Mercury 2.5 1.6 from 2 benchmarks.
- Widest category
- Math, 96.0 points apart on average over 1 shared benchmark.
- Widest single gap
- IOI (Vals AI): GPT-6 Astra 100.0% against Mercury 2.5 2.4%, 97.6 points apart on a 0 to 100 scale.
- Output price
- GPT-6 Astra $50.00 against Mercury 2.5 $0.15 per million output tokens, 333.3x apart.
- Context window
- GPT-6 Astra takes 1.05M tokens, Mercury 2.5 260K.
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 Astra | Bars | Mercury 2.5 | Ties |
|---|---|---|---|---|
| Coding | 7 | 1 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 5 | 0 | ||
| Reasoning | 1 | 0 | ||
| Math | 1 | 0 | ||
| Knowledge | 1 | 0 | ||
| General | 2 | 0 | ||
| All shared boards | 18 | 1 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index71
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| LiveBench coding | 93.7% | Not measured | Only GPT-6 Astra measured |
| LMArena WebDev | 1774 | Not measured | Only GPT-6 Astra measured |
| AA Coding Index | 76.9 | Not measured | Only GPT-6 Astra measured |
| Terminal-Bench 4 (Vals AI) | 59.6% | 0.0% | GPT-6 Astra |
| Vibe Code Bench (Vals AI) | 89.6% | 3.7% | GPT-6 Astra |
| Code migration (Vals AI) | 67.7% | 4.5% | GPT-6 Astra |
| ProgramBench (Vals AI) | 5.5% | Not measured | Only GPT-6 Astra measured |
| IOI (Vals AI) | 100.0% | 2.4% | GPT-6 Astra |
| Terminal-Bench 2.1 (Vals AI) | 87.3% | 34.5% | GPT-6 Astra |
| SWE-bench Multimodal (Vals AI) | 43.3% | Not measured | Only GPT-6 Astra measured |
| Vibe Code Bench 1-100 (Vals AI) | 27.6% | Not measured | Only GPT-6 Astra measured |
| SkillsBench (Vals AI) | Not measured | 18.1% | Only Mercury 2.5 measured |
| CyberBench (Vals AI) | 41.1% | 45.5% | Mercury 2.5 |
| SciCode | 56.5% | 38.5% | GPT-6 Astra |
| FrontierCode | 53.3% | Not measured | Only GPT-6 Astra measured |
| FrontierSWE | 65.5% | Not measured | Only GPT-6 Astra measured |
| GSO-Bench | 79.4% | Not measured | Only GPT-6 Astra measured |
| WeirdML | 93.3% | Not measured | Only GPT-6 Astra measured |
| MirrorCode | 46.7% | Not measured | Only GPT-6 Astra measured |
| ALE-Bench | 2951.3 | 301.6 | GPT-6 Astra |
| LMArena Text: coding | 1489 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: software and IT | 1474 | Not measured | Only GPT-6 Astra measured |
| LMArena WebDev: React | 1798 | Not measured | Only GPT-6 Astra measured |
| LMArena WebDev: HTML | 1775 | Not measured | Only GPT-6 Astra measured |
| LMArena WebDev: from an image | 1726 | Not measured | Only GPT-6 Astra measured |
| SciCode | 12.3% | Not measured | Only GPT-6 Astra measured |
| SWE-Bench Pro V2 | 96.9% | Not measured | Only GPT-6 Astra measured |
| SWE Atlas: Codebase QnA | 59.1% | Not measured | Only GPT-6 Astra measured |
| SWE Atlas: Refactoring | 59.1% | Not measured | Only GPT-6 Astra measured |
| SWE Atlas: Test Writing | 50.7% | Not measured | Only GPT-6 Astra measured |
| Neither model has a result on the other 13 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| AA Intelligence Index | 52.7 | 12.3 | GPT-6 Astra |
| Epoch Capabilities Index | 168.8 | Not measured | Only GPT-6 Astra measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| AA Agentic Index | 51.0 | Not measured | Only GPT-6 Astra measured |
| Design Arena: agents building web apps | 1374 | Not measured | Only GPT-6 Astra measured |
| Design Arena: agents building full-stack apps | 1357 | Not measured | Only GPT-6 Astra measured |
| Design Arena: agents building mobile apps | 1312 | Not measured | Only GPT-6 Astra measured |
| Design Arena: agents building Godot games | 1348 | Not measured | Only GPT-6 Astra measured |
| Design Arena: PowerPoint slides | 1217 | Not measured | Only GPT-6 Astra measured |
| Design Arena: spreadsheets | 1378 | Not measured | Only GPT-6 Astra measured |
| LiveBench research | 85.2% | Not measured | Only GPT-6 Astra measured |
| LiveBench browser | 84.0% | Not measured | Only GPT-6 Astra measured |
| LiveBench long agentic | 82.4% | Not measured | Only GPT-6 Astra measured |
| SREBench (Vals AI) | 56.9% | Not measured | Only GPT-6 Astra measured |
| Finance Agent (Vals AI) | 53.5% | 18.9% | GPT-6 Astra |
| Tax Agent Bench (Vals AI) | 20.7% | 2.5% | GPT-6 Astra |
| Legal Research Bench (Vals AI) | 39.4% | 4.3% | GPT-6 Astra |
| Harvey's Legal Agent Benchmark (Vals AI) | 7.1% | Not measured | Only GPT-6 Astra measured |
| Excel Modeling Benchmark (Vals AI) | 71.7% | 8.8% | GPT-6 Astra |
| BioMysteryBench (Vals AI) | 79.3% | Not measured | Only GPT-6 Astra measured |
| Terminal-Bench-Science (Vals AI) | 62.9% | 0.0% | GPT-6 Astra |
| Vals Science Index | 65.1% | Not measured | Only GPT-6 Astra measured |
| Time Horizon Index (Vals AI) | 90.5% | Not measured | Only GPT-6 Astra measured |
| CUA-bench (Vals AI) | 19.2% | Not measured | Only GPT-6 Astra measured |
| Robo-bench (Vals AI) | 23.3% | Not measured | Only GPT-6 Astra measured |
| APEX-Agents | 64.7% | Not measured | Only GPT-6 Astra measured |
| Remote Labor Index | 20.8% | Not measured | Only GPT-6 Astra measured |
| BALROG | 68.3% | Not measured | Only GPT-6 Astra measured |
| SpireClimb | 25.7% | Not measured | Only GPT-6 Astra measured |
| Vending-Bench 2 | $15,515 | Not measured | Only GPT-6 Astra measured |
| Neither model has a result on the other 15 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| LiveBench reasoning | 92.7% | Not measured | Only GPT-6 Astra measured |
| MysteryMechanism (Vals AI) | 53.1% | Not measured | Only GPT-6 Astra measured |
| GPQA Diamond (Artificial Analysis run) | 96.3% | Not measured | Only GPT-6 Astra measured |
| Humanity's Last Exam (Artificial Analysis run) | 54.7% | 11.8% | GPT-6 Astra |
| GPQA Diamond (Epoch AI run) | 95.8% | Not measured | Only GPT-6 Astra measured |
| ARC-AGI | 98.5% | Not measured | Only GPT-6 Astra measured |
| ARC-AGI-2 | 95.0% | Not measured | Only GPT-6 Astra measured |
| Chess puzzles | 74.0% | Not measured | Only GPT-6 Astra measured |
| Mystery game puzzles | 84.0% | Not measured | Only GPT-6 Astra measured |
| EBR-bench | 76.2% | Not measured | Only GPT-6 Astra measured |
| LMArena Text: hard prompts | 1463 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: expert | 1489 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: hard prompts in English | 1464 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision: diagrams | 1297 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision: homework | 1337 | Not measured | Only GPT-6 Astra measured |
| Humanity's Last Exam (Scale run) | 60.6% | Not measured | Only GPT-6 Astra measured |
| Neither model has a result on the other 7 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| OTIS Mock AIME | 100.0% | Not measured | Only GPT-6 Astra measured |
| FrontierMath, tiers 1 to 3 | 93.7% | Not measured | Only GPT-6 Astra measured |
| FrontierMath, tier 4 | 97.6% | Not measured | Only GPT-6 Astra measured |
| ProofBench | 99.0% | 3.0% | GPT-6 Astra |
| LMArena Text: math | 1465 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: mathematical work | 1484 | Not measured | Only GPT-6 Astra measured |
| MathArena | 81.9% | Not measured | Only GPT-6 Astra measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| LegalBench (Vals AI) | Not measured | 83.1% | Only Mercury 2.5 measured |
| MedCode (Vals AI) | 48.5% | 31.3% | GPT-6 Astra |
| SimpleQA Verified | 75.6% | Not measured | Only GPT-6 Astra measured |
| LMArena Text: science | 1451 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: medicine and healthcare | 1425 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: legal and government | 1462 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: business and finance | 1439 | Not measured | Only GPT-6 Astra measured |
| FACTS Benchmark Suite | 75.7% | Not measured | Only GPT-6 Astra measured |
| PRBench Finance | 47.5% | Not measured | Only GPT-6 Astra measured |
| PRBench Legal | 48.4% | Not measured | Only GPT-6 Astra measured |
| Neither model has a result on the other 10 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| Design Arena: websites | 1336 | Not measured | Only GPT-6 Astra measured |
| Design Arena: UI components | 1394 | Not measured | Only GPT-6 Astra measured |
| Design Arena: game dev | 1480 | Not measured | Only GPT-6 Astra measured |
| Design Arena: data viz | 1334 | Not measured | Only GPT-6 Astra measured |
| Design Arena: 3D | 1500 | Not measured | Only GPT-6 Astra measured |
| Design Arena: all code categories | 1379 | Not measured | Only GPT-6 Astra measured |
| Design Arena: SVG | 1430 | Not measured | Only GPT-6 Astra measured |
| Design Arena: ASCII art | 1353 | Not measured | Only GPT-6 Astra measured |
| LiveBench design | 82.0% | Not measured | Only GPT-6 Astra measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | GPT-6 Astra | Mercury 2.5 | Leader |
|---|---|---|---|
| Vals Index | 65.0% | Not measured | Only GPT-6 Astra measured |
| LiveBench data analysis | 83.0% | Not measured | Only GPT-6 Astra measured |
| Public Benefits Bench (Vals AI) | Not measured | 45.3% | Only Mercury 2.5 measured |
| SAGE (Vals AI) | 46.4% | Not measured | Only GPT-6 Astra measured |
| MedScribe (Vals AI) | 87.9% | 55.1% | GPT-6 Astra |
| Long context reasoning (AA-LCR) | 80.7% | 71.7% | GPT-6 Astra |
| DTBench | 97.3% | Not measured | Only GPT-6 Astra measured |
| LMCA | 64.4% | Not measured | Only GPT-6 Astra measured |
| Furniture assembly | 80.0% | Not measured | Only GPT-6 Astra measured |
| LMArena Text | 1441 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: instruction following | 1455 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: creative writing | 1420 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision | 1280 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: English | 1447 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: non-English | 1432 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: Chinese | 1489 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: German | 1459 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: Japanese | 1405 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: Korean | 1428 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: French | 1432 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: Spanish | 1422 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: Russian | 1438 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: Polish | 1422 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: multi-turn | 1447 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: longer queries | 1460 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: writing and language | 1437 | Not measured | Only GPT-6 Astra measured |
| LMArena Text: entertainment and media | 1427 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision: English | 1280 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision: Chinese | 1361 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision: reading text | 1292 | Not measured | Only GPT-6 Astra measured |
| LMArena Vision: creative writing | 1286 | Not measured | Only GPT-6 Astra 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 Astra | Mercury 2.5 | 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 Astra$0.035Fits in one request
Mercury 2.5$0.0001Fits in one request
Mercury 2.5 costs less
Repository review
50K input + 3K output tokens
GPT-6 Astra$0.650Fits in one request
Mercury 2.5$0.0024Fits in one request
Mercury 2.5 costs less
Agent loop
200K input + 10K output tokens
GPT-6 Astra$2.50Fits in one request
Mercury 2.5$0.0095Fits in one request
Mercury 2.5 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 Astra | Inception Mercury 2.5 |
|---|---|---|
| Maker | OpenAI | Inception |
| Weights | No public weights listed | No public weights listed |
| Released | September 4, 2026 | September 8, 2026 |
| Context window | 1.05M | 260K |
| Input $/Mtok | $10.00 | $0.04 |
| Output $/Mtok | $50.00 | $0.15 |
| Coding index | 85.1 | 1.6 |
| Independent measurements behind it | 5 | 2 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is GPT-6 Astra better than Mercury 2.5 for coding?
Pick GPT-6 Astra for coding: it scores 85.1 on the Vibe Coding Index against 1.6 for Mercury 2.5. Pick Mercury 2.5 for output tokens 333.3x cheaper ($0.15 vs $50.00 per million) and work measured by CyberBench (Vals AI), where it leads.
Which is cheaper, GPT-6 Astra or Mercury 2.5?
GPT-6 Astra costs $10.00 per million input tokens and $50.00 per million output tokens; Mercury 2.5 costs $0.04 and $0.15. Cost per solved task is known for neither of them.
Which has the bigger context window, GPT-6 Astra or Mercury 2.5?
GPT-6 Astra, with 1.05M tokens against 260K.
How many benchmarks is the GPT-6 Astra vs Mercury 2.5 comparison based on?
19 benchmarks have a result for both models as of October 22, 2026. GPT-6 Astra's coding index rests on 5 independent benchmarks and Mercury 2.5's on 2. Where only one model has a result, the page says so instead of naming a winner.