Claude Opus 5
- Input
- $5.00/M tok
- Output
- $25/M tok
- Blended
- $10/M tok · 3:1
- Speed
- 54tok/s
- Context
- 1Mtokens
- Cost per task
- ~$11/taskest.
Previous generation (superseded by Opus 5.5).
Rankings
Coding
Mixed#8 / 137rank by quality84.6±3.2quality / 1006/7 benchmarks measuredWriting & chat
Verified#4 / 162rank by quality89.1±4.3quality / 1003/5 benchmarks measuredSpeed frontierVision
Verified#13 / 98rank by quality76.6±4.8quality / 1003/6 benchmarks measuredHard reasoning
Verified#11 / 218rank by quality85.8±3.7quality / 1003/4 benchmarks measuredAgents
Mixed#9 / 187rank by quality83.5±4.6quality / 1004/7 benchmarks measured
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench Pro (Public)Lab-reported
#5 of 2379.2%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #4Claude Fable 5.181.2%
- #5Claude Opus 579.2%
- #6Intern-S2-397B68.5%
anthropic.com · 2026-07-24
DeepSWE v1.1Verified
#26 of 4562.5%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #25GLM-5.3 Flash63.4%
- #26Claude Opus 562.5%
- #27Step 5 Preview61.9%
artificialanalysis.ai · 2026-09-24
Terminal-Bench 4.0Verified
#12 of 10149.0%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #11GPT-6.1 Terra50.3%
- #12Claude Opus 549.0%
- #13GPT-6 Sol43.9%
artificialanalysis.ai · 2026-09-24
LiveCodeBenchEstimated
~#6 of 28 (estimated)~88.5%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #5Nemotron 3 Ultra89.0%
- ~#6Claude Opus 5Estimated88.5%
- #6Beam 1.0 501B A23B88.3%
No published score — estimated from related benchmarks
SciCodeVerified
#22 of 11856.4%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #21GPT-6 Astra56.5%
- #22Claude Opus 556.4%
- #23DeepSeek V4.1 Flash 101755.9%
artificialanalysis.ai · 2026-09-24
Writing & chat
IFBenchEstimated
~#6 of 140 (estimated)~81.2%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #5Qwen3.8 Flash-Next81.3%
- ~#6Claude Opus 5Estimated81.2%
- #6Qwen3.8 Flash-Next NVFP481.0%
No published score — estimated from related benchmarks
LMArena Text, Non-English (Elo)Verified
#6 of 561480 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #5Muse Spark 1.3Few votes1484
- #6Claude Opus 51480
- #6Gemini 3.1 Pro1480
lmarena.ai · 2026-10-08
#3 of 252133 Elo
eqbench.com · 2026-09-24
Vision
MMMU-ProVerified
#5 of 8784.7%
- #1Claude Opus 5.587.7%
- #2GPT-6 Astra86.9%
- #4Gemini 3.8 Flash85.6%
- #5Claude Opus 584.7%
- #6Kimi K3.1 Pro83.9%
artificialanalysis.ai · 2026-09-24
CharXiv ReasoningEstimated
~#2 of 12 (estimated)~85.1%
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~#4 of 8 (estimated)~90.6%
- #1MiniMax-M391.6%
- #2Kimi K391.1%
- #2Qwen3.8 27B91.1%
- ~#4Claude Opus 5Estimated90.6%
- #4Gemini 3.1 Pro85.3%
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondVerified
#11 of 19293.2%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #5Step 5 Preview93.5%
- #11Claude Opus 593.2%
- #12MiniMax-M392.9%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#6 of 19854.9%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #5Claude Sonnet 5.555.0%
- #6Claude Opus 554.9%
- #7GPT-6 Astra54.7%
artificialanalysis.ai · 2026-09-24
AIME (latest)Estimated
~#2 of 34 (estimated)~99.3%
No published score — estimated from related benchmarks
Agents
τ²-bench (Telecom)Estimated
~#3 of 120 (estimated)~98.2%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- ~#3Claude Opus 5Estimated98.2%
- #3Gemini 3.1 Pro95.6%
No published score — estimated from related benchmarks
OSWorld-VerifiedEstimated
~#3 of 15 (estimated)~84.0%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~#3 of 5 (estimated)~55.3%
No published score — estimated from related benchmarks
AutomationBench-AAVerified
#22 of 8456.6%
- #1Claude Fable 5.276.4%
- #2Muse Wonder Full71.3%
- #21DeepSeek V4 Pro (0813)56.7%
- #22Claude Opus 556.6%
- #23Qwen3.8 Max (0902)56.2%
artificialanalysis.ai · 2026-09-24
Not comparable across labs (1)
Kept for reference, not counted in rankings: these use a lab-specific task set, answer key or scoring, so scores can't be compared fairly across labs. Only published scores are shown.