← Back to directory
Anthropic
Claude Sonnet 4
Released May 22, 2025 · knowledge cutoff 2025-03
- Status
- Deprecated
- Location
- United States
- Type
- LLM
- Modality
- Multimodal
- Context window
- 200K
- Max output
- 64K
- Speed
- —
- Price ($/MTok in / out)
- $3 / $15
- Cost per task
- —
- Open weights
- No
Benchmarks7 of 10 reported
MMLU-Proretired
83.8%
GPQA Diamond
75.4%
SWE-bench Verified
72.7%
SWE-bench Pro
42.7%
AIME (latest)retired
70.5%
Humanity's Last Exam
9.2%
LMArena Elo
1386
Terminal-Bench 2.1
—
GDPval-AA v2
—
ARC-AGI-2
—
Strengths & weaknesses
Strengths
- Fast enough to live inside an editor loop while still handling multi-file changes
- Low-drama behavior — rarely over-reaches or rewrites what you didn't ask about
- The default that most people never found a reason to upgrade from
Weaknesses
- Loses coherence on runs measured in hours where the flagship holds
- Settles for a plausible answer sooner on genuinely hard problems
SWE-bench Pro 42.7% is Scale's public-set figure (also Table 1 of the SWE-Bench Pro paper, arXiv 2509.16941) — third-party, not Anthropic-reported.
For developers
API model strings
- anthropic
claude-sonnet-4-20250514 - aws-bedrock
anthropic.claude-sonnet-4-20250514-v1:0 - google-vertex
claude-sonnet-4@20250514
Licence
Proprietary — weights not released
Retirement
Jun 15, 2026
Lineage
No recorded predecessor