Anthropic model product
Claude Opus 4.8 is Anthropic's upgrade to Opus 4.7 and its most capable general-access model at release, with improvements across software engineering, agentic tool use, reasoning, computer use, and knowledge-work benchmarks while shipping at the same price ($5/$25 per million input/output tokens). Performance gains include SWE-Bench Verified (88.6%), SWE-Bench Pro (69.2%), Terminal-Bench 2.1 (74.6%), GPQA Diamond (93.6%), USAMO 2026 (96.7%), Humanity's Last Exam with tools (57.9%), OSWorld-Verified (83.4%), BrowseComp (84.3% single-agent, 88.5% multi-agent), MCP-Atlas (82.2%), and GDPval-AA (1890 Elo). The alignment assessment reports honesty improvements with around a four-fold drop in letting flaws in self-written code pass unremarked, a 17-fold drop relative to Sonnet 4.6 on dishonest agentic code summaries, and broadly improved adherence to Claude's constitution. The model defaults to high effort and exposes new 'extra' (xhigh) and 'max' levels for harder problems. Launches alongside Claude Code dynamic workflows (parallel subagents that plan, execute, and verify codebase-scale migrations), effort control in claude.ai and Cowork, and a Messages API extension that accepts system entries inside the messages array so harnesses can update instructions mid-task without breaking the prompt cache. Fast mode runs at 2.5× speed at $10/$50 per million input/output tokens, three times cheaper than fast mode on previous models. Available across Claude products, the Claude API as `claude-opus-4-8`, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
Updated Aug 10, 2026. Default version: Claude Opus 4.8
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version Claude Opus 4.8.
| Include | 0.9 | 1 | 31 | 100.0% | C | |
| ScreenSpot Pro | 0.9 | 1 | 24 | 100.0% | C | |
| DeepSearchQA | 0.9 | 2 | 8 | 85.7% | C | |
| SWE-bench Multilingual | 0.8 | 2 | 34 | 97.0% | C | |
| SWE-Bench Multimodal | 0.4 | 2 | 3 | 50.0% | C | |
| FrontierSWE | 0.8 | 3 | 15 | 85.7% | B | |
| OfficeQA Pro | 0.7 | 3 | 7 | 66.7% | C | |
| OSWorld-Verified | 0.8 | 3 | 22 | 90.5% | C | |
| SWE-Bench Pro | 0.7 | 3 | 44 | 95.3% | C | |
| SWE-Bench Verified | 0.9 | 3 | 104 | 98.1% | C | |
| CharXiv-R | 0.9 | 4 | 47 | 93.5% | C | |
| CyberGym | 0.8 | 4 | 11 | 70.0% | C | |
| Finance Agent v2 | 0.5 | 4 | 26 | 88.0% | B | |
| FrontierCode 1.1 | 0.5 | 4 | 15 | 78.6% | B | |
| Graphwalks parents >128k | 0.8 | 4 | 18 | 82.3% | C | |
| Toolathlon | 0.6 | 4 | 31 | 90.0% | C | |
| GPQA | 0.9 | 5 | 233 | 98.3% | C | |
| MCP Atlas | 0.8 | 5 | 30 | 86.2% | C | |
| HealthBench Professional | 0.6 | 6 | 9 | 37.5% | C | |
| Humanity's Last Exam | 0.6 | 6 | 92 | 94.5% | C | |
| LiveBench | 0.8 | 6 | 38 | 86.5% | B | |
| Finance Agent | 0.5 | 7 | 8 | 14.3% | C | |
| Terminal-Bench 2.0 | 0.7 | 7 | 49 | 87.5% | C | |
| DeepSWE 1.1 | 0.6 | 8 | 19 | 61.1% | B | |
| Graphwalks BFS >128k | 0.7 | 11 | 22 | 52.4% | C | |
| BrowseComp | 0.8 | 13 | 58 | 79.0% | C |
Preference and agent-evaluation signals from published Arena datasets.
| text style control | longer query | 9 | 1500.4 | 17,313 | N/A | |
| text style control | multi turn | 9 | 1499.8 | 6,950 | N/A | |
| document style control | overall | 10 | 1481.8 | 8,193 | N/A | |
| search factuality | overall | 10 | 1199.3 | 48,610 | N/A | |
| search style control | overall | 10 | 1195.6 | 48,610 | N/A | |
| text style control | hard prompts english | 10 | 1507.8 | 12,069 | N/A | |
| search | overall | 11 | 1204.9 | 48,610 | N/A | |
| text style control | industry business and management and financial operations | 11 | 1490.0 | 7,592 | N/A | |
| text style control | industry legal and government | 12 | 1495.9 | 3,096 | N/A | |
| text style control | expert | 13 | 1515.4 | 3,976 | N/A | |
| vision | homework | 13 | 1327.7 | 1,425 | N/A | |
| vision style control | english | 13 | 1281.4 | 4,791 | N/A | |
| vision style control | homework | 13 | 1321.7 | 1,425 | N/A | |
| text factuality | japanese | 14 | 1446.6 | 342 | N/A | |
| text style control | coding | 14 | 1526.8 | 10,721 | N/A | |
| webdev | webdev-react | 14 | 1541.6 | 6,622 | N/A | |
| text | longer query | 15 | 1481.1 | 17,313 | N/A | |
| text factuality | french | 15 | 1494.6 | 1,179 | N/A | |
| text factuality | industry mathematical | 15 | 1478.1 | 1,572 | N/A | |
| text style control | french | 15 | 1505.4 | 1,455 | N/A | |
| text style control | hard prompts | 15 | 1504.1 | 25,254 | N/A | |
| document | overall | 16 | 1468.9 | 8,193 | N/A | |
| text factuality | multi turn | 16 | 1485.6 | 6,254 | N/A | |
| text style control | industry life and physical and social science | 16 | 1499.8 | 6,198 | N/A | |
| text style control | industry medicine and healthcare | 16 | 1496.0 | 2,768 | N/A | |
| text style control | industry software and it services | 16 | 1513.7 | 15,148 | N/A | |
| text style control | instruction following | 16 | 1476.1 | 13,038 | N/A | |
| text style control | korean | 16 | 1444.4 | 615 | N/A | |
| vision | english | 16 | 1292.8 | 4,791 | N/A | |
| vision style control | chinese | 16 | 1332.8 | 566 | N/A | |
| vision style control | humor | 16 | 1270.9 | 434 | N/A | |
| webdev | overall | 16 | 1539.1 | 8,565 | N/A | |
| webdev | webdev | 16 | 1539.1 | 8,565 | N/A | |
| text factuality | chinese | 17 | 1517.6 | 1,719 | N/A | |
| text style control | chinese | 17 | 1527.6 | 2,021 | N/A | |
| text factuality | math | 18 | 1471.1 | 1,477 | N/A | |
| text factuality | spanish | 18 | 1456.8 | 909 | N/A | |
| text style control | creative writing | 18 | 1462.4 | 6,816 | N/A | |
| text style control | industry entertainment and sports and media | 18 | 1453.5 | 8,764 | N/A | |
| text factuality | expert | 19 | 1506.1 | 3,411 | N/A | |
| text factuality | longer query | 19 | 1488.2 | 17,268 | N/A | |
| text style control | industry writing and literature and language | 19 | 1467.7 | 9,714 | N/A | |
| vision | creative writing vision | 19 | 1284.9 | 641 | N/A | |
| vision | humor | 19 | 1284.4 | 434 | N/A | |
| vision style control | creative writing vision | 19 | 1272.3 | 641 | N/A | |
| text style control | russian | 20 | 1486.1 | 3,801 | N/A | |
| text | expert | 21 | 1491.6 | 3,976 | N/A | |
| text style control | spanish | 21 | 1471.4 | 1,156 | N/A | |
| vision | chinese | 21 | 1343.3 | 566 | N/A | |
| vision style control | overall | 21 | 1280.4 | 11,571 | N/A | |
| vision style control | ocr | 21 | 1292.5 | 8,366 | N/A | |
| text | multi turn | 22 | 1477.2 | 6,950 | N/A | |
| text factuality | industry legal and government | 22 | 1491.9 | 2,619 | N/A | |
| vision | overall | 22 | 1289.7 | 11,571 | N/A | |
| vision | ocr | 22 | 1298.1 | 8,366 | N/A | |
| text factuality | industry life and physical and social science | 23 | 1490.6 | 5,719 | N/A | |
| text factuality | hard prompts english | 24 | 1500.2 | 12,027 | N/A | |
| vision | diagram | 24 | 1301.1 | 3,125 | N/A | |
| text factuality | creative writing | 25 | 1449.1 | 6,227 | N/A | |
| text factuality | hard prompts | 25 | 1492.0 | 25,207 | N/A | |
| text factuality | industry entertainment and sports and media | 25 | 1442.4 | 8,556 | N/A | |
| text factuality | industry writing and literature and language | 25 | 1456.5 | 9,572 | N/A | |
| text style control | english | 25 | 1480.9 | 17,552 | N/A | |
| text style control | polish | 25 | 1477.8 | 694 | N/A | |
| vision style control | diagram | 25 | 1303.9 | 3,125 | N/A | |
| webdev | webdev-html | 25 | 1527.5 | 1,116 | N/A | |
| text | industry business and management and financial operations | 26 | 1456.4 | 7,592 | N/A | |
| text | industry legal and government | 26 | 1473.4 | 3,096 | N/A | |
| text factuality | industry business and management and financial operations | 26 | 1470.2 | 7,107 | N/A | |
| text factuality | instruction following | 26 | 1469.1 | 12,992 | N/A | |
| text factuality | russian | 26 | 1473.1 | 3,308 | N/A | |
| text | creative writing | 27 | 1445.4 | 6,816 | N/A | |
| text | french | 27 | 1476.9 | 1,455 | N/A | |
| text | instruction following | 27 | 1460.0 | 13,038 | N/A | |
| text | russian | 27 | 1465.5 | 3,801 | N/A | |
| text style control | industry mathematical | 27 | 1486.3 | 1,953 | N/A | |
| text | industry mathematical | 28 | 1477.6 | 1,953 | N/A | |
| text | industry writing and literature and language | 28 | 1453.0 | 9,714 | N/A | |
| text factuality | coding | 28 | 1519.2 | 10,676 | N/A | |
| text style control | overall | 28 | 1472.7 | 38,131 | N/A | |
| text style control | exclude ties | 28 | 1482.6 | 28,780 | N/A | |
| text style control | german | 28 | 1469.5 | 661 | N/A | |
| text | hard prompts english | 29 | 1475.6 | 12,069 | N/A | |
| text factuality | industry software and it services | 29 | 1502.1 | 15,107 | N/A | |
| text style control | non english | 29 | 1461.7 | 20,579 | N/A | |
| text | hard prompts | 30 | 1472.9 | 25,254 | N/A | |
| text | korean | 30 | 1423.3 | 615 | N/A | |
| text factuality | overall | 30 | 1462.8 | 38,059 | N/A | |
| text factuality | non english | 30 | 1452.5 | 20,539 | N/A | |
| text | industry life and physical and social science | 31 | 1469.5 | 6,198 | N/A | |
| text factuality | exclude ties | 31 | 1465.6 | 28,727 | N/A | |
| text style control | math | 31 | 1475.2 | 1,850 | N/A | |
| text | chinese | 32 | 1507.8 | 2,021 | N/A | |
| text | coding | 32 | 1484.3 | 10,721 | N/A | |
| text | english | 32 | 1458.3 | 17,552 | N/A | |
| text | industry entertainment and sports and media | 32 | 1432.5 | 8,764 | N/A | |
| text | math | 32 | 1467.9 | 1,850 | N/A | |
| text style control | japanese | 32 | 1441.7 | 422 | N/A | |
| text | spanish | 33 | 1456.0 | 1,156 | N/A | |
| text factuality | industry medicine and healthcare | 33 | 1486.5 | 2,349 | N/A |
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| Kenari | claude-opus-4-8 | global | N/A | N/A | 1M | |
| UnoRouter | claude-opus-4-8 | global | $0.425 | $2.13 | 1M | |
| Xpersona | claude-opus-4-8 | global | $1.5 | $9.25 | 200K | |
| DaoXE | claude-opus-4-8 | global | $5 | $25 | 1M | |
| LLM Gateway | claude-opus-4-8 | global | $5 | $25 | 1M | |
| OpenCode Zen | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Neon | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Pioneer | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Abacus | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Azure Cognitive Services | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Azure | claude-opus-4-8 | global | $5 | $25 | 1M | |
| AIHubMix | claude-opus-4-8 | global | $5 | $25 | 200K | |
| routing.run | claude-opus-4-8 | global | $5 | $25 | 1M | |
| FreeModel | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Requesty | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Modelis | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Anthropic | claude-opus-4-8 | global | $5 | $25 | 1M | |
| Venice AI | claude-opus-4-8 | global | $6 | $30 | 1M |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Claude Opus 4.8 | 82.9 | N/A | 1M | 128K | No | Proprietary |
A concise description based on the published model registry.
Claude Opus 4.8 is Anthropic's upgrade to Opus 4.7 and its most capable general-access model at release, with improvements across software engineering, agentic tool use, reasoning, computer use, and knowledge-work benchmarks while shipping at the same price ($5/$25 per million input/output tokens). Performance gains include SWE-Bench Verified (88.6%), SWE-Bench Pro (69.2%), Terminal-Bench 2.1 (74.6%), GPQA Diamond (93.6%), USAMO 2026 (96.7%), Humanity's Last Exam with tools (57.9%), OSWorld-Verified (83.4%), BrowseComp (84.3% single-agent, 88.5% multi-agent), MCP-Atlas (82.2%), and GDPval-AA (1890 Elo). The alignment assessment reports honesty improvements with around a four-fold drop in letting flaws in self-written code pass unremarked, a 17-fold drop relative to Sonnet 4.6 on dishonest agentic code summaries, and broadly improved adherence to Claude's constitution. The model defaults to high effort and exposes new 'extra' (xhigh) and 'max' levels for harder problems. Launches alongside Claude Code dynamic workflows (parallel subagents that plan, execute, and verify codebase-scale migrations), effort control in claude.ai and Cowork, and a Messages API extension that accepts system entries inside the messages array so harnesses can update instructions mid-task without breaking the prompt cache. Fast mode runs at 2.5× speed at $10/$50 per million input/output tokens, three times cheaper than fast mode on previous models. Available across Claude products, the Claude API as `claude-opus-4-8`, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Claude Opus 4.8.
Claude Opus 4.8's default version was released on May 28, 2026.
Claude Opus 4.8's official API price is $5 per million input tokens and $25 per million output tokens via Anthropic. The lowest tracked third-party offer starts at $0.425 input and $2.13 output via UnoRouter.
Claude Opus 4.8 is published under Anthropic in the model registry.
The default version has a 1M token context window.
No. The default version is not marked as having publicly available weights.
19 published provider offerings are linked to the default version.
Nearby ranked alternatives include Seed 2.1 Pro, GPT-5.6-Terra, Gemini 3.1 Pro.