Claude Opus 5.5 vs GPT-6 Astra
Updated October 20, 2026. 79 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick Claude Opus 5.5 for coding: it scores 92.2 on the Vibe Coding Index against 88.7 for GPT-6 Astra.
Pick GPT-6 Astra when the context window is larger, 1.05M against 1M tokens, or when what matters is Code migration (Vals AI) or IOI (Vals AI), where it leads.
- Shared results
- 79
- Claude Opus 5.5 only
- 26
- GPT-6 Astra only
- 13
- Like-for-like categories
- 8 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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, four public coding benchmarks
Claude Opus 5.5
Claude Opus 5.5 scores 92.2 against 88.7.
2 benchmarks behind each score
Intelligence
AA Intelligence Index
Claude Opus 5.5
Claude Opus 5.5 leads, 57.6 against 52.7.
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
GPT-6 Astra
GPT-6 Astra leads, 1338 against 1336.
Same board, both measured
General
Vals Index
Claude Opus 5.5
Claude Opus 5.5 leads, 70.2% against 65.0%.
Same board, both measured
Long documents
Listed context window
GPT-6 Astra
GPT-6 Astra takes 1.05M tokens against 1M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Claude Opus 5.5
Claude Opus 5.5 $0.014, GPT-6 Astra $0.035 per turn.
List prices
At a glance
- Coding index
- Claude Opus 5.5 92.2 from 2 benchmarks, GPT-6 Astra 88.7 from 2 benchmarks.
- Widest category
- Agentic work, 9.2 points apart on average over 21 shared benchmarks.
- Widest single gap
- Vending-Bench 2: Claude Opus 5.5 $9,235 against GPT-6 Astra $15,515, 39.2 points apart on a 0 to 100 scale.
- Output price
- Claude Opus 5.5 $20.00 against GPT-6 Astra $50.00 per million output tokens, 2.5x apart.
- Context window
- Claude Opus 5.5 takes 1M tokens, GPT-6 Astra 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 | Claude Opus 5.5 | Bars | GPT-6 Astra | Ties |
|---|---|---|---|---|
| Coding | 14 | 5 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 10 | 10 | 1 tie | |
| Reasoning | 3 | 7 | 1 tie | |
| Math | 1 | 2 | 1 tie | |
| Knowledge | 1 | 1 | ||
| Web design | 3 | 5 | ||
| General | 10 | 2 | ||
| All shared boards | 44 | 32 | 3 ties |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index145
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| LiveBench coding | 96.3% | 93.7% | Claude Opus 5.5 |
| LMArena WebDev | 1807 | 1774 | Claude Opus 5.5 |
| AA Coding Index | Not measured | 76.9 | Only GPT-6 Astra measured |
| Terminal-Bench 4 (Vals AI) | 65.2% | 59.6% | Claude Opus 5.5 |
| Vibe Code Bench (Vals AI) | 90.3% | 89.6% | Claude Opus 5.5 |
| Code migration (Vals AI) | 66.7% | 67.7% | GPT-6 Astra |
| ProgramBench (Vals AI) | 18.5% | 5.5% | Claude Opus 5.5 |
| IOI (Vals AI) | 95.1% | 100.0% | GPT-6 Astra |
| Terminal-Bench 2.1 (Vals AI) | 87.6% | 87.3% | Claude Opus 5.5 |
| SWE-bench Multimodal (Vals AI) | 70.2% | 43.3% | Claude Opus 5.5 |
| Vibe Code Bench 1-100 (Vals AI) | 30.4% | 27.6% | Claude Opus 5.5 |
| CyberBench (Vals AI) | 74.6% | 41.1% | Claude Opus 5.5 |
| SciCode | 66.9% | 56.5% | Claude Opus 5.5 |
| DeepSWE | Not measured | 74.1% | Only GPT-6 Astra measured |
| FrontierCode | 54.6% | 53.3% | Claude Opus 5.5 |
| FrontierSWE | 62.3% | 65.5% | GPT-6 Astra |
| GSO-Bench | Not measured | 79.4% | Only GPT-6 Astra measured |
| WeirdML | Not measured | 93.3% | Only GPT-6 Astra measured |
| MirrorCode | 77.4% | 46.7% | Claude Opus 5.5 |
| ALE-Bench | 2147.4 | 2951.3 | GPT-6 Astra |
| LMArena Text: coding | 1545 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: software and IT | 1538 | Not measured | Only Claude Opus 5.5 measured |
| LMArena WebDev: React | 1855 | 1798 | Claude Opus 5.5 |
| LMArena WebDev: HTML | 1753 | 1775 | GPT-6 Astra |
| LMArena WebDev: from an image | 1753 | 1726 | Claude Opus 5.5 |
| Neither model has a result on the other 13 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| AA Intelligence Index | 57.6 | 52.7 | Claude Opus 5.5 |
| Epoch Capabilities Index | 166.8 | 166.4 | Claude Opus 5.5 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index1010
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 51.0 | Only GPT-6 Astra measured |
| Design Arena: agents building web apps | 1318 | 1373 | GPT-6 Astra |
| Design Arena: agents building full-stack apps | Not measured | 1357 | Only GPT-6 Astra measured |
| Design Arena: agents building mobile apps | 1323 | 1308 | Claude Opus 5.5 |
| Design Arena: agents building Godot games | Not measured | 1348 | Only GPT-6 Astra measured |
| Design Arena: PowerPoint slides | Not measured | 1245 | Only GPT-6 Astra measured |
| LiveBench research | 85.0% | 85.2% | GPT-6 Astra |
| LiveBench browser | 80.0% | 84.0% | GPT-6 Astra |
| LiveBench long agentic | 85.0% | 82.4% | Claude Opus 5.5 |
| SREBench (Vals AI) | 33.6% | 56.9% | GPT-6 Astra |
| Finance Agent (Vals AI) | 58.6% | 53.5% | Claude Opus 5.5 |
| Tax Agent Bench (Vals AI) | 45.6% | 20.7% | Claude Opus 5.5 |
| Legal Research Bench (Vals AI) | 50.5% | 39.4% | Claude Opus 5.5 |
| Harvey's Legal Agent Benchmark (Vals AI) | 9.2% | 7.1% | Claude Opus 5.5 |
| Excel Modeling Benchmark (Vals AI) | 75.9% | 71.7% | Claude Opus 5.5 |
| BioMysteryBench (Vals AI) | 79.3% | 79.3% | Tie |
| Terminal-Bench-Science (Vals AI) | 47.1% | 62.9% | GPT-6 Astra |
| Vals Science Index | 58.7% | 65.1% | GPT-6 Astra |
| Time Horizon Index (Vals AI) | 91.3% | 90.5% | Claude Opus 5.5 |
| CUA-bench (Vals AI) | 14.0% | 19.2% | GPT-6 Astra |
| Robo-bench (Vals AI) | 14.6% | 23.3% | GPT-6 Astra |
| APEX-Agents | 73.5% | 64.7% | Claude Opus 5.5 |
| Remote Labor Index | 21.3% | 20.8% | Claude Opus 5.5 |
| BALROG | Not measured | 68.3% | Only GPT-6 Astra measured |
| SpireClimb | 3.3% | 25.7% | GPT-6 Astra |
| Vending-Bench 2 | $9,235 | $15,515 | GPT-6 Astra |
| LMArena Search | 1213 | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 12 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index37
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| LiveBench reasoning | 92.2% | 92.7% | GPT-6 Astra |
| MysteryMechanism (Vals AI) | 49.5% | 53.1% | GPT-6 Astra |
| GPQA Diamond (Artificial Analysis run) | Not measured | 96.3% | Only GPT-6 Astra measured |
| Humanity's Last Exam (Artificial Analysis run) | 61.4% | 54.7% | Claude Opus 5.5 |
| GPQA Diamond (Epoch AI run) | 90.6% | 95.8% | GPT-6 Astra |
| ARC-AGI | 98.5% | 98.5% | Tie |
| ARC-AGI-2 | 93.3% | 95.0% | GPT-6 Astra |
| Humanity's Last Exam | Not measured | 54.8% | Only GPT-6 Astra measured |
| Chess puzzles | 67.0% | 74.0% | GPT-6 Astra |
| Mystery game puzzles | 71.0% | 84.0% | GPT-6 Astra |
| EBR-bench | 71.4% | 76.2% | GPT-6 Astra |
| LMArena Text: hard prompts | 1535 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: expert | 1550 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: hard prompts in English | 1533 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Vision: diagrams | 1345 | 1297 | Claude Opus 5.5 |
| LMArena Vision: homework | 1362 | 1337 | Claude Opus 5.5 |
| Neither model has a result on the other 5 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index12
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| OTIS Mock AIME | 100.0% | 100.0% | Tie |
| FrontierMath, tiers 1 to 3 | 91.2% | 93.7% | GPT-6 Astra |
| FrontierMath, tier 4 | 95.0% | 97.6% | GPT-6 Astra |
| ProofBench | 100.0% | 99.0% | Claude Opus 5.5 |
| LMArena Text: math | 1508 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: mathematical work | 1506 | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| MedCode (Vals AI) | 49.8% | 48.5% | Claude Opus 5.5 |
| SimpleQA Verified | 72.2% | 75.6% | GPT-6 Astra |
| LMArena Text: science | 1522 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: medicine and healthcare | 1513 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: legal and government | 1510 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: business and finance | 1492 | Not measured | Only Claude Opus 5.5 measured |
| Neither model has a result on the other 11 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index35
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| Design Arena: websites | 1336 | 1338 | GPT-6 Astra |
| Design Arena: UI components | 1386 | 1400 | GPT-6 Astra |
| Design Arena: game dev | 1458 | 1474 | GPT-6 Astra |
| Design Arena: data viz | 1397 | 1333 | Claude Opus 5.5 |
| Design Arena: 3D | 1477 | 1505 | GPT-6 Astra |
| Design Arena: all code categories | 1390 | 1379 | Claude Opus 5.5 |
| Design Arena: SVG | 1332 | 1431 | GPT-6 Astra |
| Design Arena: ASCII art | Not measured | 1356 | Only GPT-6 Astra measured |
| LiveBench design | 86.9% | 82.0% | Claude Opus 5.5 |
GeneralLike-for-likePublic boards, shown as published, kept out of the index102
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | Leader |
|---|---|---|---|
| Vals Index | 70.2% | 65.0% | Claude Opus 5.5 |
| LiveBench data analysis | 80.3% | 83.0% | GPT-6 Astra |
| Public Benefits Bench (Vals AI) | 70.6% | Not measured | Only Claude Opus 5.5 measured |
| SAGE (Vals AI) | 45.8% | 46.4% | GPT-6 Astra |
| MedScribe (Vals AI) | 91.4% | 87.9% | Claude Opus 5.5 |
| Long context reasoning (AA-LCR) | 84.7% | 80.7% | Claude Opus 5.5 |
| DTBench | 98.9% | 97.3% | Claude Opus 5.5 |
| LMCA | 68.2% | 64.4% | Claude Opus 5.5 |
| Furniture assembly | 83.3% | 80.0% | Claude Opus 5.5 |
| LMArena Text | 1516 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: instruction following | 1536 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: creative writing | 1534 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Vision | 1321 | 1280 | Claude Opus 5.5 |
| LMArena Text: English | 1517 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: non-English | 1507 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: Chinese | 1584 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: French | 1505 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: Spanish | 1519 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: Russian | 1519 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: multi-turn | 1502 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: longer queries | 1531 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: writing and language | 1534 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Text: entertainment and media | 1511 | Not measured | Only Claude Opus 5.5 measured |
| LMArena Vision: English | 1317 | 1280 | Claude Opus 5.5 |
| LMArena Vision: Chinese | Not measured | 1361 | Only GPT-6 Astra measured |
| LMArena Vision: reading text | 1326 | 1292 | Claude Opus 5.5 |
| LMArena Vision: creative writing | 1320 | 1286 | Claude Opus 5.5 |
| Neither model has a result on the other 19 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Claude Opus 5.5 | GPT-6 Astra | 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 5.5 costs less
Repository review
50K input + 3K output tokens
Claude Opus 5.5 costs less
Agent loop
200K input + 10K output tokens
Claude Opus 5.5 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 | Anthropic Claude Opus 5.5 | OpenAI GPT-6 Astra |
|---|---|---|
| Maker | Anthropic | OpenAI |
| Weights | No public weights listed | No public weights listed |
| Released | September 22, 2026 | September 4, 2026 |
| Context window | 1M | 1.05M |
| Input $/Mtok | $4.00 | $10.00 |
| Output $/Mtok | $20.00 | $50.00 |
| Coding index | 92.2 | 88.7 |
| Independent measurements behind it | 2 | 2 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Opus 5.5 better than GPT-6 Astra for coding?
Pick Claude Opus 5.5 for coding: it scores 92.2 on the Vibe Coding Index against 88.7 for GPT-6 Astra. Pick GPT-6 Astra when the context window is larger, 1.05M against 1M tokens, or when what matters is Code migration (Vals AI) or IOI (Vals AI), where it leads.
Which is cheaper, Claude Opus 5.5 or GPT-6 Astra?
Claude Opus 5.5 costs $4.00 per million input tokens and $20.00 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. Cost per solved task is known for neither of them.
Which has the bigger context window, Claude Opus 5.5 or GPT-6 Astra?
GPT-6 Astra, with 1.05M tokens against 1M.
How many benchmarks is the Claude Opus 5.5 vs GPT-6 Astra comparison based on?
79 benchmarks have a result for both models as of October 20, 2026. Claude Opus 5.5's coding index rests on 2 independent benchmarks and GPT-6 Astra's on 2. Where only one model has a result, the page says so instead of naming a winner.