Claude Haiku 4.5 vs Claude Opus 4.7
Updated October 21, 2026. 71 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Claude Haiku 4.5 and Claude Opus 4.7 are level for coding, 49.8 against 51.1 on the Vibe Coding Index, so price and context decide.
Pick Claude Haiku 4.5 when output tokens cost 5.0x less, or when a solved task costs $0.56 against $2.95. Pick Claude Opus 4.7 when the context window is larger, 1M against 200K tokens, or when what matters is LMArena WebDev or AA Coding Index, where it leads.
- Shared results
- 71
- Claude Haiku 4.5 only
- 23
- Claude Opus 4.7 only
- 30
- Like-for-like categories
- 7 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
No call
Level: 49.8 against 51.1, inside 2 points.
3 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
Claude Opus 4.7
Claude Opus 4.7 leads, 40.7 against 16.9.
Same board, both measured
Agentic work
AA Agentic Index
Claude Opus 4.7
Claude Opus 4.7 leads, 38.6 against 8.0.
Same board, both measured
Reasoning
GPQA Diamond
Claude Opus 4.7
Claude Opus 4.7 leads, 90.2% against 72.2%.
Same board, both measured
Math
AIME
Claude Opus 4.7
Claude Opus 4.7 leads, 96.3% against 82.7%.
Same board, both measured
Knowledge
MMLU Pro
Claude Opus 4.7
Claude Opus 4.7 leads, 89.9% against 78.7%.
Same board, both measured
Web design
Design Arena: websites
No call
Only Claude Haiku 4.5 has a Design Arena: websites result, 1122.
Directional only
Long documents
Listed context window
Claude Opus 4.7
Claude Opus 4.7 takes 1M tokens against 200K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Claude Haiku 4.5
Claude Haiku 4.5 $0.0035, Claude Opus 4.7 $0.018 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Claude Haiku 4.5
Claude Haiku 4.5 $0.56, Claude Opus 4.7 $2.95.
Published attempt cost and pass rate
At a glance
- Coding index
- Claude Haiku 4.5 49.8 from 3 benchmarks, Claude Opus 4.7 51.1 from 1 benchmark.
- Widest category
- Agentic work, 33.9 points apart on average over 6 shared benchmarks.
- Widest single gap
- ARC-AGI-2: Claude Haiku 4.5 4.0% against Claude Opus 4.7 75.8%, 71.8 points apart on a 0 to 100 scale.
- Output price
- Claude Haiku 4.5 $5.00 against Claude Opus 4.7 $25.00 per million output tokens, 5.0x apart.
- Context window
- Claude Haiku 4.5 takes 200K tokens, Claude Opus 4.7 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 | Claude Haiku 4.5 | Bars | Claude Opus 4.7 | Ties |
|---|---|---|---|---|
| Coding | 0 | 13 | 1 tie | |
| Intelligence | 0 | 2 | ||
| Agentic work | 0 | 6 | ||
| Reasoning | 0 | 9 | ||
| Math | 0 | 4 | ||
| Knowledge | 0 | 12 | ||
| General | 1 | 23 | ||
| All shared boards | 1 | 69 | 1 tie |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index013
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| SWE-bench Verified | 66.6% | Not measured | Only Claude Haiku 4.5 measured |
| SWE-bench Multilingual | 64.7% | Not measured | Only Claude Haiku 4.5 measured |
| LiveBench coding | 48.1% | Not measured | Only Claude Haiku 4.5 measured |
| LMArena WebDev | 1330 | 1558 | Claude Opus 4.7 |
| AA Coding Index | 43.9 | 73.6 | Claude Opus 4.7 |
| LiveCodeBench (Vals AI) | 41.2% | 85.1% | Claude Opus 4.7 |
| SWE-bench Verified (Vals AI) | 66.6% | 82.0% | Claude Opus 4.7 |
| Vibe Code Bench (Vals AI) | 11.4% | 71.0% | Claude Opus 4.7 |
| Code migration (Vals AI) | 7.5% | 43.9% | Claude Opus 4.7 |
| ProgramBench (Vals AI) | 0.0% | 0.0% | Tie |
| Terminal-Bench 2.1 (Vals AI) | 43.8% | 68.5% | Claude Opus 4.7 |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 54.5% | Only Claude Opus 4.7 measured |
| Terminal-Bench | 35.5% | 80.2% | Claude Opus 4.7 |
| FrontierCode | Not measured | 38.5% | Only Claude Opus 4.7 measured |
| GSO-Bench | Not measured | 44.1% | Only Claude Opus 4.7 measured |
| WeirdML | 45.4% | 76.4% | Claude Opus 4.7 |
| MirrorCode | Not measured | 31.1% | Only Claude Opus 4.7 measured |
| LMArena Text: coding | 1454 | 1519 | Claude Opus 4.7 |
| LMArena Text: software and IT | 1439 | 1516 | Claude Opus 4.7 |
| LMArena WebDev: React | 1326 | 1560 | Claude Opus 4.7 |
| LMArena WebDev: HTML | 1318 | 1554 | Claude Opus 4.7 |
| LMArena WebDev: from an image | Not measured | 1577 | Only Claude Opus 4.7 measured |
| Neither model has a result on the other 16 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| AA Intelligence Index | 16.9 | 40.7 | Claude Opus 4.7 |
| Epoch Capabilities Index | 142.2 | 156.8 | Claude Opus 4.7 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index06
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| AA Agentic Index | 8.0 | 38.6 | Claude Opus 4.7 |
| LiveBench research | 42.2% | Not measured | Only Claude Haiku 4.5 measured |
| LiveBench browser | 40.0% | Not measured | Only Claude Haiku 4.5 measured |
| LiveBench long agentic | 11.1% | Not measured | Only Claude Haiku 4.5 measured |
| Finance Agent (Vals AI) | 31.0% | 51.5% | Claude Opus 4.7 |
| Tax Agent Bench (Vals AI) | 5.9% | 23.7% | Claude Opus 4.7 |
| Legal Research Bench (Vals AI) | 10.6% | 38.5% | Claude Opus 4.7 |
| Excel Modeling Benchmark (Vals AI) | 22.7% | 63.7% | Claude Opus 4.7 |
| τ²-Bench Telecom | Not measured | 88.6% | Only Claude Opus 4.7 measured |
| APEX-Agents | Not measured | 49.2% | Only Claude Opus 4.7 measured |
| OSWorld 2.0 | Not measured | 18.2% | Only Claude Opus 4.7 measured |
| BALROG | 31.2% | Not measured | Only Claude Haiku 4.5 measured |
| DeepResearch Bench | 45.5% | Not measured | Only Claude Haiku 4.5 measured |
| PostTrainBench | Not measured | 28.6% | Only Claude Opus 4.7 measured |
| Vending-Bench 2 | $459 | $10,937 | Claude Opus 4.7 |
| LMArena Search | Not measured | 1230 | Only Claude Opus 4.7 measured |
| Neither model has a result on the other 23 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index09
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| GPQA Diamond | 72.2% | 90.2% | Claude Opus 4.7 |
| LiveBench reasoning | 61.7% | Not measured | Only Claude Haiku 4.5 measured |
| MMMU Pro (Vals AI) | 46.1% | 85.5% | Claude Opus 4.7 |
| GPQA Diamond (Artificial Analysis run) | Not measured | 91.4% | Only Claude Opus 4.7 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 42.3% | Only Claude Opus 4.7 measured |
| GPQA Diamond (Epoch AI run) | 71.2% | 90.2% | Claude Opus 4.7 |
| ARC-AGI | 47.7% | 93.5% | Claude Opus 4.7 |
| ARC-AGI-2 | 4.0% | 75.8% | Claude Opus 4.7 |
| SimpleBench | Not measured | 61.7% | Only Claude Opus 4.7 measured |
| Chess puzzles | 8.0% | 41.0% | Claude Opus 4.7 |
| Mystery game puzzles | Not measured | 28.0% | Only Claude Opus 4.7 measured |
| EBR-bench | Not measured | 19.0% | Only Claude Opus 4.7 measured |
| LMArena Text: hard prompts | 1421 | 1506 | Claude Opus 4.7 |
| LMArena Text: expert | 1443 | 1521 | Claude Opus 4.7 |
| LMArena Text: hard prompts in English | 1435 | 1508 | Claude Opus 4.7 |
| LMArena Vision: diagrams | Not measured | 1336 | Only Claude Opus 4.7 measured |
| LMArena Vision: homework | Not measured | 1337 | Only Claude Opus 4.7 measured |
| 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 Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| AIME | 82.7% | 96.3% | Claude Opus 4.7 |
| MGSM (Vals AI) | 92.2% | Not measured | Only Claude Haiku 4.5 measured |
| OTIS Mock AIME | 66.7% | 97.8% | Claude Opus 4.7 |
| FrontierMath, tiers 1 to 3 | Not measured | 70.2% | Only Claude Opus 4.7 measured |
| FrontierMath, tier 4 | Not measured | 31.7% | Only Claude Opus 4.7 measured |
| ProofBench | Not measured | 54.0% | Only Claude Opus 4.7 measured |
| MATH level 5 | 96.4% | Not measured | Only Claude Haiku 4.5 measured |
| LMArena Text: math | 1398 | 1498 | Claude Opus 4.7 |
| LMArena Text: mathematical work | 1430 | 1509 | Claude Opus 4.7 |
| Neither model has a result on the other 4 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index012
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| MMLU Pro | 78.7% | 89.9% | Claude Opus 4.7 |
| LegalBench (Vals AI) | 81.2% | 85.3% | Claude Opus 4.7 |
| CaseLaw (Vals AI) | 56.5% | 68.4% | Claude Opus 4.7 |
| CorpFin (Vals AI) | 60.6% | 66.1% | Claude Opus 4.7 |
| TaxEval (Vals AI) | 67.5% | 75.3% | Claude Opus 4.7 |
| MortgageTax (Vals AI) | 62.2% | 70.3% | Claude Opus 4.7 |
| MedQA (Vals AI) | 79.6% | Not measured | Only Claude Haiku 4.5 measured |
| MedCode (Vals AI) | 32.7% | 54.9% | Claude Opus 4.7 |
| SimpleQA Verified | 13.2% | 51.7% | Claude Opus 4.7 |
| LMArena Text: science | 1406 | 1508 | Claude Opus 4.7 |
| LMArena Text: medicine and healthcare | 1394 | 1495 | Claude Opus 4.7 |
| LMArena Text: legal and government | 1400 | 1504 | Claude Opus 4.7 |
| LMArena Text: business and finance | 1400 | 1490 | Claude Opus 4.7 |
| Neither model has a result on the other 4 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| Design Arena: websites | 1122 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: UI components | 1100 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: game dev | 1097 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: data viz | 1124 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: 3D | 1072 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: all code categories | 1115 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: SVG | 1025 | Not measured | Only Claude Haiku 4.5 measured |
| Design Arena: ASCII art | 1139 | Not measured | Only Claude Haiku 4.5 measured |
| LiveBench design | 59.9% | Not measured | Only Claude Haiku 4.5 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index123
| Benchmark | Claude Haiku 4.5 | Claude Opus 4.7 | Leader |
|---|---|---|---|
| LiveBench data analysis | 59.3% | Not measured | Only Claude Haiku 4.5 measured |
| Public Benefits Bench (Vals AI) | 54.3% | Not measured | Only Claude Haiku 4.5 measured |
| SAGE (Vals AI) | 31.8% | 56.1% | Claude Opus 4.7 |
| MedScribe (Vals AI) | 85.2% | 83.0% | Claude Haiku 4.5 |
| Vals Multimodal Index | 42.9% | 67.4% | Claude Opus 4.7 |
| IFBench | Not measured | 58.6% | Only Claude Opus 4.7 measured |
| Long context reasoning (AA-LCR) | Not measured | 78.7% | Only Claude Opus 4.7 measured |
| DTBench | 73.6% | 94.7% | Claude Opus 4.7 |
| LMCA | 30.9% | 52.2% | Claude Opus 4.7 |
| Furniture assembly | Not measured | 33.3% | Only Claude Opus 4.7 measured |
| ExploitBench | 13.7% | 26.5% | Claude Opus 4.7 |
| LMArena Text | 1397 | 1490 | Claude Opus 4.7 |
| LMArena Text: instruction following | 1414 | 1498 | Claude Opus 4.7 |
| LMArena Text: creative writing | 1371 | 1486 | Claude Opus 4.7 |
| LMArena Vision | Not measured | 1317 | Only Claude Opus 4.7 measured |
| LMArena Text: English | 1414 | 1496 | Claude Opus 4.7 |
| LMArena Text: non-English | 1377 | 1480 | Claude Opus 4.7 |
| LMArena Text: Chinese | 1418 | 1531 | Claude Opus 4.7 |
| LMArena Text: German | 1374 | 1495 | Claude Opus 4.7 |
| LMArena Text: Japanese | 1338 | 1474 | Claude Opus 4.7 |
| LMArena Text: Korean | 1346 | 1464 | Claude Opus 4.7 |
| LMArena Text: French | 1411 | 1503 | Claude Opus 4.7 |
| LMArena Text: Spanish | 1421 | 1494 | Claude Opus 4.7 |
| LMArena Text: Russian | 1381 | 1494 | Claude Opus 4.7 |
| LMArena Text: Polish | 1343 | 1489 | Claude Opus 4.7 |
| LMArena Text: multi-turn | 1409 | 1505 | Claude Opus 4.7 |
| LMArena Text: longer queries | 1427 | 1505 | Claude Opus 4.7 |
| LMArena Text: writing and language | 1386 | 1492 | Claude Opus 4.7 |
| LMArena Text: entertainment and media | 1369 | 1474 | Claude Opus 4.7 |
| LMArena Vision: English | Not measured | 1322 | Only Claude Opus 4.7 measured |
| LMArena Vision: Chinese | Not measured | 1371 | Only Claude Opus 4.7 measured |
| LMArena Vision: reading text | Not measured | 1327 | Only Claude Opus 4.7 measured |
| LMArena Vision: humor | Not measured | 1309 | Only Claude Opus 4.7 measured |
| LMArena Vision: recognizing things | Not measured | 1278 | Only Claude Opus 4.7 measured |
| LMArena Vision: creative writing | Not measured | 1326 | Only Claude Opus 4.7 measured |
| LMArena Document | 1420 | 1495 | Claude Opus 4.7 |
| 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 | Claude Haiku 4.5 | Claude Opus 4.7 | 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 Haiku 4.5 costs less
Repository review
50K input + 3K output tokens
Claude Haiku 4.5 costs less
Agent loop
200K input + 10K output tokens
Claude Haiku 4.5 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 Haiku 4.5 | Anthropic Claude Opus 4.7 |
|---|---|---|
| Maker | Anthropic | Anthropic |
| Weights | No public weights listed | No public weights listed |
| Released | October 15, 2025 | April 16, 2026 |
| Context window | 200K | 1M |
| Input $/Mtok | $1.00 | $5.00 |
| Output $/Mtok | $5.00 | $25.00 |
| Coding index | 49.8 | 51.1 |
| Independent measurements behind it | 3 | 1 |
| Cost per solved task | $0.56 | $2.95 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Haiku 4.5 better than Claude Opus 4.7 for coding?
Claude Haiku 4.5 and Claude Opus 4.7 are level for coding, 49.8 against 51.1 on the Vibe Coding Index, so price and context decide. Pick Claude Haiku 4.5 when output tokens cost 5.0x less, or when a solved task costs $0.56 against $2.95. Pick Claude Opus 4.7 when the context window is larger, 1M against 200K tokens, or when what matters is LMArena WebDev or AA Coding Index, where it leads.
Which is cheaper, Claude Haiku 4.5 or Claude Opus 4.7?
Claude Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens; Claude Opus 4.7 costs $5.00 and $25.00. Per solved SWE-bench Verified task, Claude Haiku 4.5 costs $0.56 and Claude Opus 4.7 $2.95.
Which has the bigger context window, Claude Haiku 4.5 or Claude Opus 4.7?
Claude Opus 4.7, with 1M tokens against 200K.
How many benchmarks is the Claude Haiku 4.5 vs Claude Opus 4.7 comparison based on?
71 benchmarks have a result for both models as of October 21, 2026. Claude Haiku 4.5's coding index rests on 3 independent benchmarks and Claude Opus 4.7's on 1. Where only one model has a result, the page says so instead of naming a winner.