Skip to main content
BenchLM
Advertise

Model profile

Claude Opus 4.8

AnthropicCurrentReleased May 28, 2026
Data verified
Overall Score
78.34Public #5 of 200Verified #5 of 99
Arena Elo
1485
Eligible category ranks
5of 8
Price (1M tokens)
$5 in / $25 out
API pricing
Speed
Not listed
Context
1M

Evidence coverage

53 of 323 tracked benchmarks are published. 31 are verified and 22 provisional. 8 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
53 / 323
Verified
31
Provisional
22
Categories with evidence
8 / 8

Evidence by category

  • Agentic21 benchmarks
    Mixed evidence
  • Coding10 benchmarks
    Mixed evidence
  • Reasoning2 benchmarks
    Reported
  • Knowledge10 benchmarks
    Mixed evidence
  • Math3 benchmarks
    Verified
  • Multilingual1 benchmark
    Verified
  • Multimodal5 benchmarks
    Mixed evidence
  • Inst. Following1 benchmark
    Reported
ProprietaryReasoning
Confidence:
High
base

Claude Opus 4.8 ranks #5 out of 200 models on the public leaderboard with an overall score of 78.34/100. It also ranks #5 out of 99 on the verified leaderboard. This places it among the top tier of AI models available in 2026, competing directly with the strongest models from leading AI labs.

Claude Opus 4.8 is a proprietary model with a 1M token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

BenchLM links it directly to Claude Opus 4.7 (Adaptive) as the earlier related model in that lineage. This profile currently has 53 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Multimodal & Grounded (#1), while its weakest is Agentic (#8). This performance profile makes it particularly strong for screenshots, documents, charts, and grounded multimodal workflows.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 73.51–81.96

  1. GPT-5.6 Sol
    OpenAI
    #381.96
    GPT-5.6 Sol is #3 with a score of 81.96.
    Compare
  2. Kimi K3
    Moonshot AI
    #480.96
    Kimi K3 is #4 with a score of 80.96.
    Compare
  3. Claude Opus 4.8Current model
    Anthropic
    #578.34
    Claude Opus 4.8 is #5 with a score of 78.34.
  4. Muse Spark 1.1
    Meta
    #677.44
    Muse Spark 1.1 is #6 with a score of 77.44.
    Compare
  5. Grok 4.5
    xAI
    #776.72
    Grok 4.5 is #7 with a score of 76.72.
    Compare
  6. GPT-5.4
    OpenAI
    #874.24
    GPT-5.4 is #8 with a score of 74.24.
    Compare
  7. GPT-5.5
    OpenAI
    #973.51
    GPT-5.5 is #9 with a score of 73.51.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Multimodal100%
    Eligible cohort rank #1 of 29Category score 91.5
  2. Knowledge98%
    Eligible cohort rank #2 of 52Category score 92.5
  3. Math83%
    Eligible cohort rank #2 of 7Category score 66.9
  4. Coding95%
    Eligible cohort rank #7 of 122Category score 70.6
  5. Agentic94%
    Eligible cohort rank #8 of 119Category score 64.0

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #8 of 119Percentile 94thWeight 22%21 benchmarksMixed sources64.0
CodingRank #7 of 122Percentile 95thWeight 20%10 benchmarksMixed sources70.6
ReasoningWeight 17%2 benchmarksReportedScore pending
KnowledgeRank #2 of 52Percentile 98thWeight 12%10 benchmarksMixed sources92.5
MathRank #2 of 7Percentile 83rdWeight 5%3 benchmarksVerified66.9
MultilingualWeight 7%1 benchmarkVerifiedScore pending
MultimodalRank #1 of 29Percentile 100thWeight 12%5 benchmarksMixed sources91.5
Inst. FollowingWeight 5%1 benchmarkReportedScore pending

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1485±5.130,024
Coding1535±7.88,366
Math1496±15.51,471
Instruction Following1494±7.310,141
Creative Writing1464±9.35,301
Multi-turn1506±9.15,367
Hard Prompts1514±6.219,745
Hard Prompts (English)1514±7.49,594
Longer Query1506±7.013,426

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic21 benchmarks
Terminal-Bench 2.0Provider exact
74.6%Weighted 38%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.3 reports Terminal-Bench 2.1 at 74.6% mean reward using the Terminus-2 harness.
OSWorld-VerifiedProvider exact
83.4%Weighted 34%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.12.6 reports OSWorld-Verified at 83.4% first-attempt success rate.
BrowseCompProvider exact
84.3%Weighted 28%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.10.2 reports single-agent BrowseComp at 84.3%.
DeepSearchQAProvider exact
93.1%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.10.3 reports DeepSearchQA F1 at 93.1%.
Finance Agent v2Provider exact
53.9%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.13.2 reports Finance Agent Benchmark v2 at 53.92%; BenchLM rounds display to one decimal.
GDPval-AAProvider exact
1594Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.13.6 reports GDPval-AA at 1890 Elo.
MCP AtlasProvider exact
82.2%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.13.4 reports MCP-Atlas pass rate at 82.2%.
ToolathlonProvider exact
59.9%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.13.7 reports Toolathlon Pass@1 at 59.9%.
Gert LabsBenchmark exact

Gert Labs Composite Game Benchmark

72.97%Display only
Source: Gert Labs rankingsProvenance: Gert Labs reports this composite leaderboard score in the public rankings API. BenchLM scales the source gscore from 0-1 to 0-100 and stores it as a display-only agentic benchmark.
AA Agentic IndexReported

Artificial Analysis Agentic Index

47.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
τ²-bench resultsReported

τ²-Bench Tool-Agent-User Evaluation

94.4%Display only
Source: Artificial Analysis: tau2-bench leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
GDPval-AAReported

GDPval-AA normalized

54.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
ResearchClawBenchBenchmark exact
21.1%Display only
Source: ResearchClawBench leaderboardProvenance: ResearchClawBench reports this model as ResearchHarness (Claude-Opus-4.8) in the official Pass@1 leaderboard. BenchLM stores the one-decimal RADS average on the local ResearchClawBench display key and excludes it from weighted rankings.
OSWorld 2.0Benchmark exact
20.6%Display only
Source: OSWorld 2.0 paperProvenance: OSWorld 2.0 reports Claude Opus 4.8 batched actions on its 500-step main table. BenchLM stores the binary completion score and notes the corresponding partial score was 54.8%.
AA BriefcaseReported

Artificial Analysis Briefcase

1347Display only
Source: Artificial Analysis: aa-briefcase leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA AutomationBenchReported

Artificial Analysis AutomationBench

48.5%Display only
Source: Artificial Analysis: automationbench-aa leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA EnterpriseOps-GymReported

Artificial Analysis EnterpriseOps-Gym

44.0%Display only
Source: Artificial Analysis: enterprise-ops-gym-aa leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA Harvey LABReported

Artificial Analysis Harvey LAB-AA

91.1%Display only
Source: Artificial Analysis: harvey-lab-aa leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA Tau3 BankingReported

Artificial Analysis Tau3-Banking

27.6%Display only
Source: Artificial Analysis: tau3-banking leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
terminalBenchHardReported
58.3%Display only
Source: Artificial Analysis: terminalbench-hard leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
aaTerminalBench21Reported
84.6%Display only
Source: Artificial Analysis: terminalbench-v2-1 leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Coding10 benchmarks
SWE-bench VerifiedProvider exact

Software Engineering Benchmark Verified

88.6%Weighted 16%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.2 reports SWE-bench Verified at 88.6%.
SWE-bench ProProvider exact
69.2%Weighted 10%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.2 reports SWE-bench Pro at 69.2%.
SWE MultilingualProvider exact
84.4%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.2 reports SWE-bench Multilingual at 84.4%.
SWE MultimodalProvider exact

SWE-bench Multimodal

38.4%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.2 reports SWE-bench Multimodal at 38.4%.
Terminal-Bench 2.0Provider exact
74.6%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.3 reports Terminal-Bench 2.1 at 74.6% mean reward using the Terminus-2 harness.
cursorBench31Benchmark exact
58.4%Display only
Source: Cursor evals: CursorBench 3.1Provenance: Cursor reports Opus 4.8 High at this exact CursorBench 3.1 score on its public evals page. BenchLM stores it on the Claude Opus 4.8 row as a display-only coding-agent benchmark.
cursorBench32Benchmark exact
62.3%Display only
Source: Cursor evals: CursorBench 3.2Provenance: Cursor reports Opus 4.8 Max at this exact CursorBench 3.2 score on its public evals page. BenchLM stores it on the Claude Opus 4.8 row as a display-only coding-agent benchmark.
AA Coding IndexReported

Artificial Analysis Coding Index

74.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-SciCodeReported

Artificial Analysis SciCode

53.5%Display only
Source: Artificial Analysis: scicode leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
FrontierCode 1.1 MainBenchmark exact
46.5%Display only
Source: Cognition: FrontierCode 1.1Provenance: Cognition reports Claude Opus 4.8 at 46.5% on FrontierCode 1.1 Main, using the best max effort row from the published data JSON.
Reasoning2 benchmarks
AA-LCRReported

Artificial Analysis Long Context Reasoning

67.7%Display only
Source: Artificial Analysis: artificial-analysis-long-context-reasoning leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

20.9%Display only
Source: Artificial Analysis: critpt leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
Knowledge10 benchmarks
HLEProvider exact

Humanity's Last Exam

57.9%Weighted 45%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.10.1 reports Humanity's Last Exam with tools at 57.9%.
GPQAProvider exact

Graduate-Level Google-Proof Q&A

93.6%Weighted 7%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.6 reports GPQA Diamond at 93.6%; BenchLM stores that exact value on both gpqaDiamond and the weighted gpqa lane.
GPQA-DProvider exact

GPQA Diamond

93.6%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.6 reports GPQA Diamond at 93.6%.
HLE w/o toolsProvider exact

Humanity's Last Exam without tools

49.8%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.10.1 reports Humanity's Last Exam without tools at 49.8%.
Artificial Analysis Intelligence IndexReported
55.7%Display only
Source: Artificial Analysis: artificial-analysis-intelligence-index leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

92.0%Display only
Source: Artificial Analysis: gpqa-diamond leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

45.7%Display only
Source: Artificial Analysis: humanitys-last-exam leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

27.4%Display only
Source: Artificial Analysis: omniscience leaderboardProvenance: Display-only row synced from the current Artificial Analysis evaluation leaderboard. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

46.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

35.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Math3 benchmarks
FrontierMath v2 (Tiers 1-3)Benchmark exact

FrontierMath v2 Tiers 1-3

47.241%Weighted 30%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tiers 1-3 at 47.241% for claude-opus-4-8_max. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
USAMO 2026Provider exact

United States of America Mathematical Olympiad 2026

96.7%Weighted 10%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.7 reports USAMO 2026 at 96.7%.
FrontierMath v2 (Tier 4)Benchmark exact

FrontierMath v2 Tier 4

31.250%Weighted 10%
Source: Epoch AI FrontierMath v2 leaderboardProvenance: Epoch AI reports FrontierMath v2 Tier 4 at 31.25% for claude-opus-4-8_max. BenchLM selects the highest published thinking effort for the model and stores the v2 benchmark slice separately.
Multilingual1 benchmark
INCLUDEProvider exact
87.6%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.15.3 reports INCLUDE average accuracy at 87.6%.
Multimodal5 benchmarks
OfficeQA ProProvider exact
66.2%Weighted 30%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.13.1 reports OfficeQA Pro at 66.2% exact-match accuracy under Anthropic's internal agentic harness.
CharXivProvider exact

CharXiv Reasoning

89.9%Weighted 25%
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.12.4 reports CharXiv Reasoning with Python tools at 89.9%.
ScreenSpot ProProvider exact
87.9%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.12.5 reports ScreenSpot-Pro with Python tools at 87.9%.
CharXiv w/o toolsProvider exact

CharXiv Reasoning without tools

80.5%Display only
Source: Anthropic: Claude Opus 4.8 system cardProvenance: Section 8.12.4 reports CharXiv Reasoning without tools at 80.5%.
Design Arena WebsiteReported

Design Arena Website Elo

1270Display only
Source: OpenRouter model benchmarksProvenance: Display-only Design Arena Website Elo synced from OpenRouter model benchmark metadata. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

62.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.

Claude Opus 4.8 Family

Base entry

Related Earlier Model

Claude Opus 4.7 (Adaptive)

Frequently Asked Questions

How does Claude Opus 4.8 perform overall in AI benchmarks?

Claude Opus 4.8 currently ranks #5 out of 200 models on BenchLM's provisional leaderboard with an overall score of 78.34. It also ranks #5 out of 99 on the verified leaderboard. It is created by Anthropic. Its published context window is 1M.

Is Claude Opus 4.8 good for knowledge and understanding?

Claude Opus 4.8 ranks #2 out of 52 models in knowledge and understanding benchmarks with an average score of 92.5. It is among the top performers in this category.

Is Claude Opus 4.8 good for coding and programming?

Claude Opus 4.8 ranks #7 out of 122 models in coding and programming benchmarks with an average score of 70.6. It is among the top performers in this category.

Is Claude Opus 4.8 good for mathematics?

Claude Opus 4.8 ranks #2 out of 7 models in mathematics benchmarks with an average score of 66.9. It is among the top performers in this category.

Is Claude Opus 4.8 good for reasoning and logic?

Claude Opus 4.8 has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is Claude Opus 4.8 good for agentic tool use and computer tasks?

Claude Opus 4.8 ranks #8 out of 119 models in agentic tool use and computer tasks benchmarks with an average score of 64. It is among the top performers in this category.

Is Claude Opus 4.8 good for multimodal and grounded tasks?

Claude Opus 4.8 ranks #1 out of 29 models in multimodal and grounded tasks benchmarks with an average score of 91.5. It is among the top performers in this category.

Is Claude Opus 4.8 good for instruction following?

Claude Opus 4.8 has visible benchmark coverage in instruction following, but BenchLM does not currently assign it a global category rank there.

Is Claude Opus 4.8 good for multilingual tasks?

Claude Opus 4.8 has visible benchmark coverage in multilingual tasks, but BenchLM does not currently assign it a global category rank there.

Does Claude Opus 4.8 have full benchmark coverage on BenchLM?

Not yet. Claude Opus 4.8 currently has 53 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of Claude Opus 4.8?

Claude Opus 4.8 has a published context window of 1M, which determines how much text it can process in a single interaction.

Related Resources

More from Anthropic LLM Pricing Comparison BenchLM Methodology The AI Race Which LLM Should I Use? Benchmark Explainers
Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.