← Back to directory
Anthropic
Claude 3 Opus
Released Mar 4, 2024 · knowledge cutoff 2023-08
- Status
- Deprecated
- Location
- United States
- Modality
- Multimodal
- Context window
- 200K
- Max output
- 4K
- Speed
- —
- Price ($/MTok in / out)
- $15 / $75
- Cost per task
- —
- Open weights
- No
Benchmarks4 of 8 reported
MMLU-Pro
68.5%
GPQA Diamond
50.4%
SWE-bench Verified
22.7%
LMArena Elo
1263
Terminal-Bench 2.1
—
AIME (latest)
—
Humanity's Last Exam
—
ARC-AGI-2
—
Strengths & weaknesses
Strengths
- Prose quality still cited as a high-water mark — varied rhythm, few assistant tics
- Reads subtext, tone and intent where its contemporaries flattened everything
- Keeps a long document genuinely in view and answers across it without losing the thread
Weaknesses
- Slow enough that even trivial requests felt deliberate
- Over-cautious — hedges or refuses on benign medical, legal and safety-adjacent questions
- One-pass answers, so hard maths and logic fail with no recovery
Flagship of the Claude 3 generation.
For developers
API model strings
Not researched
Licence
Proprietary — weights not released
Retirement
Retirement date not researched
Lineage
No recorded predecessor