Inference360
Anthropic

Claude Opus 4.8

Claude Opus 4.8 is a reasoning model from Anthropic. 58 benchmarks count toward its score, in 7 categories.

availableShows if the model has enough results for an index.
IndexOverall score out of 100.71.1 ±2.6
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.58/s
Input / 1MUS dollars per 1M input tokens.$5 batch $2.5
Output / 1MUS dollars per 1M output tokens.$25 batch $12.5 US dollars per 1M output tokens in a batch.
ContextMaximum tokens in one request.1M
EloLMArena rating and rank.1453 (#40)

The index is a score out of 100. The ± range shows how much it can change. Batch work costs less.

53,446 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category, out of 100.

Out of 100
AgenticMulti-step tasks with tools.
69.7
CodingCode writing and repair.
71.6
ReasoningLogic problems and puzzles.
70.4
MultimodalTasks with images and text.
71.3
KnowledgeFacts and expert knowledge.
69.6
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
63.0
MathMath problems.
74.0

Results

58 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result as a score out of 100.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
OTIS Mock AIME 2024-2025Math98.3%66.5max effort—Epoch AI
United States of America Mathematical Olympiad 2026Math96.7%———Mathematical Association of America
τ²-Bench Tool-Agent-User EvaluationAgentic94.4%66.5——Victor Barres et al.
LiveBench MathematicsMath94.3%73.9max effort25 Jun 2026LiveBench
Graduate-Level Google-Proof Q&AKnowledge93.6%64.7——David Rein et al.
GPQA DiamondKnowledge93.6%64.7——David Rein et al.
DeepSearchQAAgentic93.1%72.1——Meta AI
ARC-AGI-1 (semi-private)Reasoning92.5%70.9max effort—ARC Prize Foundation
GPQA DiamondKnowledge92.4%63.6—1 Sept 2026Vals AI
Artificial Analysis GPQA DiamondKnowledge92.0%62.8——Artificial Analysis
GPQA diamondKnowledge91.0%62.3max effort—Epoch AI
CharXiv ReasoningMultimodal89.9%68.5——CharXiv authors
MMLU ProKnowledge89.6%61.7—1 Sept 2026Vals AI
LiveBench ReasoningReasoning89.2%79.9max effort25 Jun 2026LiveBench
Software Engineering Benchmark VerifiedCoding88.6%68.6——Carlos E. Jimenez et al.
SWE-benchCoding88.6%68.6—1 Sept 2026Vals AI
ScreenSpot ProMultimodal87.9%70.1——Kaixin Li et al.
LiveCodeBenchCoding87.8%64.2—1 Sept 2026Vals AI
INCLUDEMultilingual87.6%———Qwen
MMMU ProMultimodal86.6%68.3—1 Sept 2026Vals AI
BrowseCompAgentic84.3%70.6——OpenAI
OSWorld-VerifiedAgentic83.4%71.4——Tianbao Xie et al.
Vibe Code Bench v1.1Coding82.7%76.6OpenHands21 Sept 2026Vals AI
MCP AtlasAgentic82.2%70.0——OpenAI
LiveBench CodingCoding81.8%73.5max effort25 Jun 2026LiveBench
CharXiv Reasoning without toolsMultimodal80.5%———CharXiv authors
FrontierMath-Tiers-1-3-v2-PrivateMath80.0%76.5max effort—Epoch AI
LiveBench LanguageKnowledge79.7%66.7max effort25 Jun 2026LiveBench
Artificial Analysis Long Context ReasoningReasoning77.7%62.0——Artificial Analysis
Artificial Analysis Coding IndexCoding74.3%71.3——Artificial Analysis
Gert Labs Composite Game BenchmarkAgentic73.0%77.1——Gert Labs
ARC-AGI-2 (semi-private)Reasoning72.1%76.1high effort—ARC Prize Foundation
LiveBench Instruction FollowingInstruction72.0%73.7max effort25 Jun 2026LiveBench
Terminal-Bench 2.1Agentic71.9%66.5—21 Sept 2026Vals AI
Terminal-Bench 2.0Agentic70.0%72.7—4 Jun 2026Vals AI
SWE-bench ProCoding69.2%70.9——Xiang Deng et al.
OfficeQA ProMultimodal66.2%78.5——OfficeQA Pro authors
LiveBench Data AnalysisReasoning66.0%47.7max effort25 Jun 2026LiveBench
cursorBench32Coding62.3%70.7——Benchmark authors
Artificial Analysis IFBenchInstruction62.2%52.2——Artificial Analysis
ToolathlonAgentic59.9%73.8——OpenAI
SkillsBenchCoding59.2%73.2OpenHands11 Sept 2026Vals AI
cursorBench31Coding58.4%———Benchmark authors
Humanity's Last ExamKnowledge57.9%77.8——Center for AI Safety et al.
FrontierMath-Tier-4-v2-PrivateMath56.1%74.9max effort—Epoch AI
Artificial Analysis SciCodeCoding54.4%68.4——Artificial Analysis
SimpleQA VerifiedKnowledge53.0%70.6max effort—Epoch AI
LiveBench Agentic CodingAgentic50.5%64.2max effort25 Jun 2026LiveBench
Humanity's Last Exam without toolsKnowledge49.8%71.0——OpenAI
Artificial Analysis Omniscience AccuracyKnowledge48.8%74.2——Artificial Analysis
Artificial Analysis Humanity's Last ExamKnowledge48.7%77.4——Artificial Analysis
Code MigrationCoding47.3%75.2—21 Sept 2026Vals AI
FrontierMath-2025-02-28-PrivateMath47.2%74.7max effort—Epoch AI
GDPval-AA normalizedAgentic46.9%71.6——Artificial Analysis
FrontierCode 1.1 MainCoding46.5%74.9——Cognition
Artificial Analysis AnalystAgentAgentic45.0%73.1——Artificial Analysis
Artificial Analysis Agentic IndexAgentic42.6%71.9——Artificial Analysis
Furniture AssemblyReasoning42.5%69.8max effort—Epoch AI
Artificial Analysis Intelligence IndexKnowledge41.8%74.7——Artificial Analysis
τ²-bench BankingAgentic39.7%27.3max effort · Sierra4 Aug 2026Sierra Research
OEIS Open LiteMath39.0%—high effort—Epoch AI
Mystery Game PuzzlesReasoning36.0%72.2max effort—Epoch AI
Chess PuzzlesReasoning34.0%67.5max effort—Epoch AI
FrontierMath-Tier-4-2025-07-01-PrivateMath31.3%77.4max effort—Epoch AI
OEIS OpenMath29.9%—high effort—Epoch AI
EBR-benchReasoning28.6%68.9max effort—Epoch AI
ResearchClawBenchAgentic21.1%———InternScience
Terminal-Bench 3.0Agentic21.1%70.9——Ryan Marten et al.
Critical Physics TasksReasoning20.9%82.2——Artificial Analysis
OSWorld 2.0Agentic20.6%65.8——Mengqi Yuan et al.
Terminal-Bench 4.0Agentic16.2%70.9—21 Sept 2026Vals AI
Agent Arena steerabilityAgentic11.179.8high effort15 Sept 2026LMArena
Agent Arena command recoveryAgentic7.575.8high effort15 Sept 2026LMArena
Agent Arena task outcomeAgentic6.374.5high effort15 Sept 2026LMArena
ARC-AGI-3 (semi-private)Reasoning1.5%—high effort—ARC Prize Foundation
ProgramBenchCoding1.0%——21 Sept 2026Vals AI

58 benchmarks count, from 67 of 76 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsOpenRouter, collected directlyNo licence statedEpoch AI, collected directlyCC BY — free to use and redistribute with attributionLiveBench, collected directlyNo licence stated for the leaderboard. The site repo serving these CSVs has no LICENSE; the harness repo carries upstream Apache-2.0 and MIT copies that cover code, not resultsARC Prize Foundation, collected directlyNo licence stated. Their terms ask for written permission before commercial useVals AI, collected directlyNo licence stated. Read from the public leaderboard and credited to Vals AISierra Research, collected directlyMIT — results are in the licensed repositoryLMArena, collected directlyCC BY 4.0 (lmarena-ai/leaderboard-dataset on Hugging Face)

Same level, lower price

GPT-6 Luna68.4 · $0.5DeepSeek V4.1 Flash70.6 · $0.6MiMo-V2.6-Pro73.5 · $0.87

More from Anthropic

Claude Opus 5.582.8Claude Fable 5.181.4Claude Opus 577.9Claude Fable 577.5Claude Sonnet 567.6