Inference
EN
Anthropic

Claude Opus 4.8

Claude Opus 4.8 is a reasoning model from Anthropic. 67 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Reasoning
  • Input: text · image · file
  • Released 2026-05
RankedEnough results for an index.
IndexAll categories, out of 100.67.7±2.0Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.69/sMedian 67/s
First tokenSeconds to the first output token.0.8 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$5.00batch $2.50Median $0.60
Output / 1MUS dollars per 1M output tokens.$25.00batch $12.50 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1MMedian 262K
EloLMArena rating and rank.1454±4 #48Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
anthropic/claude-opus-4.8
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.50
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
8 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Amazon BedrockCloud$5.00$25.001M
DigitalOceanCloud$5.00$25.001M
GMI CloudHost$5.00$25.001M
Google Vertex AICloud$5.00$25.001M
Microsoft AzureCloud$5.00$25.001M
OpenRouterRouter$5.00$25.001M
SAP AI CoreCloud$5.00$25.001M
Snowflake CortexCloud$5.00$25.001M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.21 counted66.538 / 218
CodingCode writing and repair.14 counted68.326 / 209
ReasoningLogic problems and puzzles.10 counted68.031 / 227
MultimodalTasks with images and text.7 counted66.521 / 84
KnowledgeFacts and expert knowledge.6 counted66.436 / 268
MultilingualTasks in many languages.1 counted60.3—
InstructionTasks with strict prompt rules.2 counted59.714 / 63
MathMath problems.6 counted71.219 / 168

Results

67 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Gert Labs Composite Game BenchmarkAgentic73.0%1 / 52
Terminal-Bench 2.0 · Vals AIAgentic70.0% ±1.42 / 66
Software Engineering Benchmark VerifiedCoding88.6%5 / 114
Graduate-Level Google-Proof Q&AKnowledge93.6%8 / 141
LiveBench Browser UseAgenticAbove fitFar above this model's index.96.0%3 / 49
MMLU ProKnowledge89.6% ±0.39 / 135
Artificial Analysis Humanity's Last ExamKnowledge48.7%15 / 222
FrontierMath-2025-02-28-PrivateMath47.2% ±2.95 / 69
LiveCodeBenchCoding87.8% ±0.911 / 142
ToolathlonAgentic59.9%2 / 22
FrontierMath-Tier-4-2025-07-01-PrivateMath31.3% ±6.85 / 55
GPQA Diamond · BenchLMKnowledge93.6%8 / 84
ScreenSpot ProMultimodal87.9%2 / 20
SWE-bench ProCoding69.2%9 / 83
Artificial Analysis GPQA DiamondKnowledge92.0%22 / 199
Humanity's Last ExamKnowledge57.9%9 / 77
GPQA Diamond · Vals AIKnowledge92.4% ±2.016 / 135
Critical Physics TasksReasoning20.9%26 / 214
OSWorld-VerifiedAgentic83.4%5 / 37
SWE-benchCoding88.6% ±1.412 / 87
OTIS Mock AIME 2024-2025Math98.3% ±1.428 / 200
GPQA diamond · Epoch AIKnowledge91.0% ±1.933 / 217
Artificial Analysis Omniscience AccuracyKnowledge48.8%25 / 164
DeepSearchQAAgentic93.1%4 / 24
INCLUDEMultilingual87.6%2 / 12
LiveBench CodingCoding81.8%11 / 66
τ²-Bench Tool-Agent-User EvaluationAgentic94.4%25 / 149
Terminal-Bench HardAgentic58.3%4 / 23
Agent Arena command recoveryAgentic7.29 / 50
Humanity's Last Exam without toolsKnowledge49.8%10 / 53
Chess PuzzlesReasoning52.0% ±5.017 / 90
Vibe Code Bench v1.1Coding82.7% ±3.124 / 123
MMMU ProMultimodal86.6% ±0.818 / 92
CharXiv ReasoningMultimodal89.9%10 / 49
MirrorCodeCoding65.6% ±10.88 / 39
ARC-AGI-2 (semi-private)Reasoning72.1%27 / 131
Code MigrationCoding47.3% ±4.216 / 77
ARC-AGI-1 (semi-private)Reasoning92.5%28 / 126
FrontierMath-Tiers-1-3-v2-PrivateMath80.0% ±2.420 / 90
MCP AtlasAgentic82.2%12 / 52
LiveBench ReasoningReasoning89.2%19 / 80
GDPval-AA normalizedAgentic47.7%32 / 133
LiveBench MathematicsMath94.3%16 / 66
SkillsBenchCoding59.2% ±4.59 / 35
LiveBench Long AgenticAgentic48.3%14 / 51
SimpleQA VerifiedKnowledge53.0% ±1.625 / 91
Mystery Game PuzzlesReasoning36.0% ±4.822 / 80
LiveBench Instruction FollowingInstruction72.0%19 / 66
Artificial Analysis SciCodeCoding54.4%33 / 114
FrontierMath-Tier-4-v2-PrivateMath56.1% ±7.821 / 72
LiveBench ResearchAgentic81.8%16 / 52
LiveBench DesignAgentic78.8%16 / 52
BrowseCompAgentic84.3%20 / 63
τ²-bench BankingAgentic39.7%9 / 28
Terminal-Bench 2.1 · Vals AIAgentic71.9% ±0.625 / 75
Artificial Analysis Long Context ReasoningReasoning77.7%74 / 217
FrontierCode 1.1 MainCoding46.5%6 / 17
OfficeQA ProMultimodal66.2%6 / 16
Terminal-Bench 3.0Agentic21.1%5 / 13
Artificial Analysis IFBenchInstruction62.2%61 / 146
LiveBench LanguageKnowledge79.7%31 / 66
Artificial Analysis Terminal-Bench v2.1Agentic84.6%10 / 21
MazeBench: visual spatial reasoning in a 3D open worldAgentic0.0%23 / 48
Furniture AssemblyReasoning42.5% ±6.418 / 37
cursorBench32Coding62.3%9 / 18
ARC-AGI-3 (semi-private)Reasoning1.5%13 / 26
Terminal-Bench 4.0Agentic23.2% ±1.331 / 61
Terminal-Bench 2.1 · BenchLMAgentic74.6%45 / 86
PostTrainBench v1.1Coding32.9%8 / 15
EBR-benchReasoning28.6% ±1.417 / 31
OSWorld 2.0Agentic20.6%15 / 25
Terminal-Bench 4.0.0Agentic23.6% ±1.814 / 23
Terminal-Bench 4.1.0Agentic23.6% ±1.814 / 23
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal86.8%16 / 26
LiveBench Agentic CodingAgentic50.5%41 / 66
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal62.3%25 / 39
LiveBench Coding 2.0Coding77.3%34 / 52
Artificial Analysis AnalystAgentAgentic45.0%17 / 26
CharXiv Reasoning without toolsMultimodal80.5%13 / 17
LiveBench Data AnalysisReasoning66.0%63 / 80
FrontierCode 2.0Coding14.7%26 / 27
MortgageTaxFinance69.9% ±0.95 / 97
MedCodeHealthcare53.2% ±2.210 / 106
TaxEval v2Finance75.6% ±0.814 / 143
CorpFin v2Finance66.7% ±0.916 / 131
Public Benefits Bench v1Public benefits62.1% ±1.32 / 13
EMBFinance69.4% ±2.615 / 75
MedScribeHealthcare85.8% ±1.926 / 109
ProgramBenchCoding1.0% ±0.716 / 65
Public Benefits Bench v1.1Public benefits68.1% ±1.213 / 50
Legal Research BenchLegal43.8% ±3.425 / 90
OEIS OpenMath29.9% ±2.11 / 3
LegalBenchLegal83.6% ±0.551 / 149
Finance Agent (v2)Finance53.9% ±0.229 / 79
Tax Agent BenchFinance30.0% ±3.334 / 82
Terminal-Bench-ScienceScience4.3% ±2.426 / 44
OEIS Open LiteMath39.0% ±4.93 / 5
Time Horizon Index: KSPOther11.8%8 / 13
United States of America Mathematical Olympiad 2026Math96.7%—
cursorBench31Coding58.4%—
Artificial Analysis EnterpriseOps-GymAgentic44.0%—
ResearchClawBenchAgentic21.1%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYLiveBenchNo licence statedARC Prize FoundationTerms restrict reuseVals AINo licence statedSierra ResearchMITTerminal-BenchNo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index69.0 → 67.7
23 Oct 202612 Nov 2026
Output / 1M$25.00 → $25.00
23 Oct 202612 Nov 2026

Questions

How much does Claude Opus 4.8 cost?
$5.00 per 1M input tokens, $25.00 per 1M output tokens.
How good is Claude Opus 4.8?
Its index is 67.7 out of 100. That is place 32 of 283 ranked models.
What is Claude Opus 4.8 best at?
Math: 71.2 out of 100, place 19 of 168.
How fast is Claude Opus 4.8?
69 output tokens per second, 0.8 s to the first token.
What is the context window of Claude Opus 4.8?
1M tokens.
Is Claude Opus 4.8 open source?
No. Its weights are not published.
When was Claude Opus 4.8 released?
May 2026.

Badge

Claude Opus 4.8 on Inference360: place 32 of 283, index 67.7
HTML
<a href="https://inference360.ai/models/claude-opus-4-8"><img src="https://inference360.ai/badges/claude-opus-4-8.svg" alt="Claude Opus 4.8 on Inference360: place 32 of 283, index 67.7" height="20"></a>
Markdown
[![Claude Opus 4.8 on Inference360: place 32 of 283, index 67.7](https://inference360.ai/badges/claude-opus-4-8.svg)](https://inference360.ai/models/claude-opus-4-8)

Compare with

Grok 4.667.8Gemini 3.8 Flash67.8Ling 3.1 Flash67.2Kimi K366.9Claude Opus 4.762.0

Lower price

Ling 3.1 Flash67.2 · $0.135/1MDeepSeek V4.1 Flash 101768.5 · $0.40/1MClaude Haiku 5.566.3 · $0.50/1M

More from Anthropic

All models
Claude Fable 574.0Claude Mythos 570.8Claude Fable LatestUnrankedClaude Haiku LatestUnrankedClaude Sonnet LatestUnranked