Claude Opus 5 vs GPT-6.2 Astra
GPT-6.2 Astra leads Claude Opus 5 on coding, 87.0 to 81.7 on the Vibe Coding Index, and Claude Opus 5 costs 1.6x less per output token ($25.00 vs $40.00 per million). Updated November 22, 2026; 64 benchmarks have a result for both.
Claude Opus 5 or GPT-6.2 Astra: which to pick
Pick GPT-6.2 Astra for coding: it scores 87.0 on the Vibe Coding Index against 81.7 for Claude Opus 5, though on 4 benchmarks against 5.
Pick Claude Opus 5 for output tokens 1.6x cheaper ($25.00 vs $40.00 per million) and work measured by MirrorCode or LMArena Text: coding, where it leads.
- Shared results
- 64
- Claude Opus 5 only
- 78
- GPT-6.2 Astra only
- 12
- Like-for-like categories
- 8 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-6.2 Astra
GPT-6.2 Astra scores 87.0 against 81.7.
5 and 4 benchmarks behind the scores
Intelligence
AA Intelligence Index
GPT-6.2 Astra
GPT-6.2 Astra leads, 62.1 against 44.2.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Claude Opus 5 has a GPQA Diamond result, 93.4%.
Directional only
Knowledge
MMLU Pro
No call
Only Claude Opus 5 has a MMLU Pro result, 91.6%.
Directional only
Web design
Design Arena: websites
GPT-6.2 Astra
GPT-6.2 Astra leads, 1333 against 1299.
Same board, both measured
General
Vals Index
No call
Only GPT-6.2 Astra has a Vals Index result, 67.7%.
Directional only
Long documents
Listed context window
GPT-6.2 Astra
GPT-6.2 Astra takes 1.05M tokens against 1M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Claude Opus 5
Claude Opus 5 $0.018, GPT-6.2 Astra $0.028 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Claude Opus 5 has the number, $1.33.
Directional only
At a glance
- Coding index
- Claude Opus 5 81.7 from 5 benchmarks, GPT-6.2 Astra 87.0 from 4 benchmarks.
- Widest category
- Reasoning, 19.1 points apart on average over 14 shared benchmarks.
- Widest single gap
- EBR-bench: Claude Opus 5 45.7% against GPT-6.2 Astra 92.4%, 46.7 points apart on a 0 to 100 scale.
- Output price
- Claude Opus 5 $25.00 against GPT-6.2 Astra $40.00 per million output tokens, 1.6x apart.
- Context window
- Claude Opus 5 takes 1M tokens, GPT-6.2 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 | Bars | GPT-6.2 Astra | Ties |
|---|---|---|---|---|
| Coding | 3 | 8 | ||
| Intelligence | 0 | 2 | ||
| Agentic work | 4 | 8 | ||
| Reasoning | 3 | 11 | ||
| Math | 0 | 4 | ||
| Knowledge | 3 | 2 | ||
| Web design | 0 | 5 | ||
| General | 10 | 1 | ||
| All shared boards | 23 | 41 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index38
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 91.9% | Only GPT-6.2 Astra measured |
| LMArena WebDev | 1692 | 1766 | GPT-6.2 Astra |
| LiveCodeBench (Vals AI) | 89.0% | Not measured | Only Claude Opus 5 measured |
| SWE-bench Verified (Vals AI) | 97.0% | Not measured | Only Claude Opus 5 measured |
| Terminal-Bench 4 (Vals AI) | 53.5% | 63.1% | GPT-6.2 Astra |
| Vibe Code Bench (Vals AI) | 88.4% | 89.9% | GPT-6.2 Astra |
| Code migration (Vals AI) | 57.5% | Not measured | Only Claude Opus 5 measured |
| ProgramBench (Vals AI) | 3.0% | Not measured | Only Claude Opus 5 measured |
| IOI (Vals AI) | 84.3% | Not measured | Only Claude Opus 5 measured |
| Terminal-Bench 2.1 (Vals AI) | 84.6% | Not measured | Only Claude Opus 5 measured |
| Vibe Code Bench 1-100 (Vals AI) | 28.5% | Not measured | Only Claude Opus 5 measured |
| SkillsBench (Vals AI) | 60.4% | Not measured | Only Claude Opus 5 measured |
| CyberBench (Vals AI) | 65.4% | Not measured | Only Claude Opus 5 measured |
| SciCode | 56.4% | Not measured | Only Claude Opus 5 measured |
| Terminal-Bench 4.0 (Artificial Analysis run) | 49.0% | 64.1% | GPT-6.2 Astra |
| FrontierCode 2.0 | 31.2% | 49.0% | GPT-6.2 Astra |
| FrontierCode | 53.4% | Not measured | Only Claude Opus 5 measured |
| FrontierSWE | 52.0% | Not measured | Only Claude Opus 5 measured |
| WeirdML | 91.8% | Not measured | Only Claude Opus 5 measured |
| MirrorCode | 87.8% | 67.8% | Claude Opus 5 |
| ALE-Bench | 2164.6 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: coding | 1532 | 1450 | Claude Opus 5 |
| LMArena Text: software and IT | 1526 | 1443 | Claude Opus 5 |
| LMArena WebDev: React | 1713 | 1770 | GPT-6.2 Astra |
| LMArena WebDev: HTML | 1681 | 1854 | GPT-6.2 Astra |
| LMArena WebDev: from an image | 1664 | Not measured | Only Claude Opus 5 measured |
| SciCode | 13.8% | Not measured | Only Claude Opus 5 measured |
| SWE-Bench Pro V2 | 99.4% | Not measured | Only Claude Opus 5 measured |
| SWE Atlas: Codebase QnA | 63.2% | Not measured | Only Claude Opus 5 measured |
| SWE Atlas: Test Writing | 62.2% | Not measured | Only Claude Opus 5 measured |
| GameBench 2 | 84.8 | 97.2 | GPT-6.2 Astra |
| Terminal-Bench 3.0 | 42.7% | Not measured | Only Claude Opus 5 measured |
| Terminal-Bench 4.0 | 53.9% | Not measured | Only Claude Opus 5 measured |
| ProgramBench | 4.5% | Not measured | Only Claude Opus 5 measured |
| Neither model has a result on the other 18 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| AA Intelligence Index | 44.2 | 62.1 | GPT-6.2 Astra |
| Epoch Capabilities Index | 164.7 | 172.8 | GPT-6.2 Astra |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index48
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| Design Arena: agents building web apps | 1236 | 1334 | GPT-6.2 Astra |
| Design Arena: agents building full-stack apps | 1285 | 1351 | GPT-6.2 Astra |
| Design Arena: agents building mobile apps | 1348 | 1305 | Claude Opus 5 |
| Design Arena: agents building Android apps | 1243 | 1257 | GPT-6.2 Astra |
| Design Arena: agents building Godot games | 1249 | Not measured | Only Claude Opus 5 measured |
| Design Arena: agents building games | 1203 | Not measured | Only Claude Opus 5 measured |
| Design Arena: PowerPoint slides | 1225 | Not measured | Only Claude Opus 5 measured |
| Design Arena: spreadsheets | 1241 | 1278 | GPT-6.2 Astra |
| LiveBench research | Not measured | 91.5% | Only GPT-6.2 Astra measured |
| LiveBench browser | Not measured | 93.0% | Only GPT-6.2 Astra measured |
| LiveBench long agentic | Not measured | 70.8% | Only GPT-6.2 Astra measured |
| SREBench (Vals AI) | 12.2% | Not measured | Only Claude Opus 5 measured |
| Finance Agent (Vals AI) | 58.6% | Not measured | Only Claude Opus 5 measured |
| Tax Agent Bench (Vals AI) | 47.1% | 21.8% | Claude Opus 5 |
| Legal Research Bench (Vals AI) | 55.3% | 42.8% | Claude Opus 5 |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 16.3% | Only GPT-6.2 Astra measured |
| Excel Modeling Benchmark (Vals AI) | 73.6% | Not measured | Only Claude Opus 5 measured |
| BioMysteryBench (Vals AI) | 79.3% | 80.7% | GPT-6.2 Astra |
| Terminal-Bench-Science (Vals AI) | 27.1% | 71.4% | GPT-6.2 Astra |
| Vals Science Index | 47.9% | 68.4% | GPT-6.2 Astra |
| Time Horizon Index (Vals AI) | 18.8% | Not measured | Only Claude Opus 5 measured |
| CUA-bench (Vals AI) | 9.0% | Not measured | Only Claude Opus 5 measured |
| Robo-bench (Vals AI) | Not measured | 32.7% | Only GPT-6.2 Astra measured |
| APEX-Agents | 65.8% | 72.2% | GPT-6.2 Astra |
| OSWorld 2.0 | 31.4% | Not measured | Only Claude Opus 5 measured |
| BALROG | 63.4% | Not measured | Only Claude Opus 5 measured |
| PostTrainBench | 35.7% | Not measured | Only Claude Opus 5 measured |
| SpireClimb | Not measured | 41.3% | Only GPT-6.2 Astra measured |
| Vending-Bench 2 | $11,182 | Not measured | Only Claude Opus 5 measured |
| LMArena Search | 1238 | 1191 | Claude Opus 5 |
| MCP Atlas | 85.8% | Not measured | Only Claude Opus 5 measured |
| τ³-Banking | 48.7% | Not measured | Only Claude Opus 5 measured |
| Neither model has a result on the other 12 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index311
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| GPQA Diamond | 93.4% | Not measured | Only Claude Opus 5 measured |
| LiveBench reasoning | Not measured | 91.6% | Only GPT-6.2 Astra measured |
| MysteryMechanism (Vals AI) | 37.4% | 53.1% | GPT-6.2 Astra |
| MMMU Pro (Vals AI) | 89.9% | Not measured | Only Claude Opus 5 measured |
| GPQA Diamond (Artificial Analysis run) | 93.7% | Not measured | Only Claude Opus 5 measured |
| Humanity's Last Exam (Artificial Analysis run) | 54.9% | Not measured | Only Claude Opus 5 measured |
| HLE-Diamond (Artificial Analysis run) | 44.8% | 72.2% | GPT-6.2 Astra |
| CritPt v2, hard subset (Artificial Analysis run) | 50.0% | 71.5% | GPT-6.2 Astra |
| GPQA Diamond (Epoch AI run) | 93.9% | 96.8% | GPT-6.2 Astra |
| ARC-AGI | 97.5% | 98.5% | GPT-6.2 Astra |
| ARC-AGI-2 | 90.4% | 95.8% | GPT-6.2 Astra |
| SimpleBench | 80.6% | 77.3% | Claude Opus 5 |
| Chess puzzles | 43.0% | 78.0% | GPT-6.2 Astra |
| Mystery game puzzles | 59.0% | 95.0% | GPT-6.2 Astra |
| EBR-bench | 45.7% | 92.4% | GPT-6.2 Astra |
| LMArena Text: hard prompts | 1525 | 1425 | Claude Opus 5 |
| LMArena Text: expert | 1553 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: hard prompts in English | 1523 | 1426 | Claude Opus 5 |
| LMArena Vision: diagrams | 1345 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: homework | 1336 | Not measured | Only Claude Opus 5 measured |
| Humanity's Last Exam (Scale run) | 38.6% | 66.9% | GPT-6.2 Astra |
| GoBench | 2254 | 2493 | GPT-6.2 Astra |
| Neither model has a result on the other 4 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| OTIS Mock AIME | 98.9% | 100.0% | GPT-6.2 Astra |
| FrontierMath, tiers 1 to 3 | 85.6% | 92.6% | GPT-6.2 Astra |
| FrontierMath, tier 4 | 73.2% | 100.0% | GPT-6.2 Astra |
| ProofBench | 99.0% | Not measured | Only Claude Opus 5 measured |
| LMArena Text: math | 1532 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: mathematical work | 1543 | Not measured | Only Claude Opus 5 measured |
| MathArena | 64.1% | 92.9% | GPT-6.2 Astra |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index32
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| MMLU Pro | 91.6% | Not measured | Only Claude Opus 5 measured |
| LegalBench (Vals AI) | 87.0% | Not measured | Only Claude Opus 5 measured |
| CorpFin (Vals AI) | 73.2% | Not measured | Only Claude Opus 5 measured |
| TaxEval (Vals AI) | 75.1% | Not measured | Only Claude Opus 5 measured |
| MortgageTax (Vals AI) | 72.1% | Not measured | Only Claude Opus 5 measured |
| MedCode (Vals AI) | 63.6% | Not measured | Only Claude Opus 5 measured |
| AA-Omniscience | 37.6 | 50.0 | GPT-6.2 Astra |
| SimpleQA Verified | 59.9% | 78.5% | GPT-6.2 Astra |
| LMArena Text: science | 1533 | 1415 | Claude Opus 5 |
| LMArena Text: medicine and healthcare | 1526 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: legal and government | 1527 | 1415 | Claude Opus 5 |
| LMArena Text: business and finance | 1499 | 1413 | Claude Opus 5 |
| Neither model has a result on the other 9 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| Design Arena: websites | 1299 | 1333 | GPT-6.2 Astra |
| Design Arena: UI components | 1335 | 1341 | GPT-6.2 Astra |
| Design Arena: game dev | 1316 | 1501 | GPT-6.2 Astra |
| Design Arena: data viz | 1324 | Not measured | Only Claude Opus 5 measured |
| Design Arena: 3D | 1309 | 1430 | GPT-6.2 Astra |
| Design Arena: all code categories | 1310 | 1389 | GPT-6.2 Astra |
| Design Arena: SVG | 1298 | Not measured | Only Claude Opus 5 measured |
| Design Arena: ASCII art | 1338 | Not measured | Only Claude Opus 5 measured |
| LiveBench design | Not measured | 82.8% | Only GPT-6.2 Astra measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index101
| Benchmark | Claude Opus 5 | GPT-6.2 Astra | Leader |
|---|---|---|---|
| Vals Index | Not measured | 67.7% | Only GPT-6.2 Astra measured |
| LiveBench data analysis | Not measured | 84.1% | Only GPT-6.2 Astra measured |
| Public Benefits Bench (Vals AI) | 76.9% | Not measured | Only Claude Opus 5 measured |
| MedScribe (Vals AI) | 91.0% | Not measured | Only Claude Opus 5 measured |
| Vals Multimodal Index | 73.9% | Not measured | Only Claude Opus 5 measured |
| Long context reasoning (AA-LCR) | 82.0% | Not measured | Only Claude Opus 5 measured |
| GDPval-AA | 1729 | Not measured | Only Claude Opus 5 measured |
| DTBench | 97.9% | Not measured | Only Claude Opus 5 measured |
| LMCA | 64.5% | Not measured | Only Claude Opus 5 measured |
| Furniture assembly | 60.8% | 89.2% | GPT-6.2 Astra |
| LMArena Text | 1506 | 1414 | Claude Opus 5 |
| LMArena Text: instruction following | 1517 | 1423 | Claude Opus 5 |
| LMArena Text: creative writing | 1494 | 1417 | Claude Opus 5 |
| LMArena Vision | 1317 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: English | 1507 | 1417 | Claude Opus 5 |
| LMArena Text: non-English | 1500 | 1404 | Claude Opus 5 |
| LMArena Text: Chinese | 1575 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: German | 1519 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: Japanese | 1522 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: Korean | 1503 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: French | 1521 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: Spanish | 1522 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: Russian | 1507 | 1437 | Claude Opus 5 |
| LMArena Text: Polish | 1523 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: multi-turn | 1499 | 1425 | Claude Opus 5 |
| LMArena Text: longer queries | 1516 | 1431 | Claude Opus 5 |
| LMArena Text: writing and language | 1499 | 1440 | Claude Opus 5 |
| LMArena Text: entertainment and media | 1480 | 1413 | Claude Opus 5 |
| LMArena Vision: English | 1310 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: Chinese | 1401 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: reading text | 1325 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: humor | 1293 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: creative writing | 1308 | Not measured | Only Claude Opus 5 measured |
| LMArena Document | 1509 | Not measured | Only Claude Opus 5 measured |
| MultiChallenge | Not measured | 73.2% | Only GPT-6.2 Astra measured |
| EQ-Bench 4 | 1365 | Not measured | Only Claude Opus 5 measured |
| Neither model has a result on the other 13 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Claude Opus 5 | GPT-6.2 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 costs less
Repository review
50K input + 3K output tokens
Claude Opus 5 costs less
Agent loop
200K input + 10K output tokens
Claude Opus 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 | OpenAI GPT-6.2 Astra |
|---|---|---|
| Maker | Anthropic | OpenAI |
| Weights | No public weights listed | No public weights listed |
| Released | July 24, 2026 | November 17, 2026 |
| Context window | 1M | 1.05M |
| Input $/Mtok | $5.00 | $8.00 |
| Output $/Mtok | $25.00 | $40.00 |
| Coding index | 81.7 | 87.0 |
| Independent measurements behind it | 5 | 4 |
| Cost per solved task | $1.33 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Opus 5 better than GPT-6.2 Astra for coding?
Pick GPT-6.2 Astra for coding: it scores 87.0 on the Vibe Coding Index against 81.7 for Claude Opus 5, though on 4 benchmarks against 5. Pick Claude Opus 5 for output tokens 1.6x cheaper ($25.00 vs $40.00 per million) and work measured by MirrorCode or LMArena Text: coding, where it leads.
Which is cheaper, Claude Opus 5 or GPT-6.2 Astra?
Claude Opus 5 costs $5.00 per million input tokens and $25.00 per million output tokens; GPT-6.2 Astra costs $8.00 and $40.00. Cost per solved task is known for Claude Opus 5 only.
Which has the bigger context window, Claude Opus 5 or GPT-6.2 Astra?
GPT-6.2 Astra, with 1.05M tokens against 1M.
How many benchmarks is the Claude Opus 5 vs GPT-6.2 Astra comparison based on?
64 benchmarks have a result for both models as of November 22, 2026. Claude Opus 5's coding index rests on 5 independent benchmarks and GPT-6.2 Astra's on 4. Where only one model has a result, the page says so instead of naming a winner.