Chaptered summary is still being generated for this document. Showing a heuristic brief in the meantime.
Summary
34,762-word document condensed to 175 words. Anthropic · Aug 3, 2026
TL;DR
“This system card describes Claude Sonnet 5, the latest model in Anthropic’s Sonnet family. It is an upgrade to Claude Sonnet 4.6, with gains in various aspects of agentic performance.”
Top benchmarks
| Benchmark | Variant | Score |
|---|---|---|
| DNA Synthesis Screening Evasion | — | 5.0% |
| Long-form virology tasks | task_1 | 82.0% |
| Long-form virology tasks | task_1 | 79.0% |
| VCT | — | 56.0% |
| VCT | — | 50.0% |
| VCT | — | 47.0% |
| VCT | — | 45.0% |
| VCT | — | 22.1% |
Showing top 8 of 58. See full list below.
Safety findings
- “We believe these mitigations make catastrophic risk in this category low but still not negligible, for reasons discussed in our most recent Risk Report .”
- “We believe these risk mitigations are equal to or stronger than our historical ASL-3 protections and sufficient to make catastrophic risk in this category very low but not negligible (further discussion of our reasoning can be found in our most recent Risk Report) .”
- “We believe Sonnet 5 does not change the picture presented for this threat model in our most recent Risk Report .”
Deployment scope
- “available through other widely used models and tools.”
What’s new
- •“Added a footnote in 8.9.2 BrowseComp with a reproduction recipe via public Messages API. 2”
Every italicized passage is a verbatim substring of the source document (checked deterministically after extraction). Field selection is heuristic — some quotes may lack surrounding context and some claims may be absent if no matching pattern appeared. For citation, open the source: original model card · source SHA 7b6f526a27ff · version dated Aug 3, 2026.
Extracted Evaluations(58 results)
Sort by:0/58 rows fully reproducible (0%)
| Benchmark | Category | State | Score | Setup | Source |
|---|---|---|---|---|---|
| agent | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
/ verified | agent | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| agent | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
/ pro | coding | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
/ multilingual | coding | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
/ multimodal | coding | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
/ verified | coding | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| medical | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
/ professional | medical | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| other | scored | 5.0 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
Long-form virology tasks/ task_1 | other | scored | 0.8 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Long-form virology tasks/ task_1 | other | scored | 0.8 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| other | scored | 0.6 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
| other | scored | 0.5 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
| other | scored | 0.5 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
| other | scored | 0.5 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
| other | scored | 0.2 | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
BenchCAD | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Terminal-Bench/ 2.1 | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
FrontierCode | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
CursorBench | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
USAMO/ 2026 | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
ArxivMath | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
ProgramBench | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
GDP.pdf | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
ChartMuseum | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
CharXiv/ reasoning | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
/ v2 | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Legal Agent Benchmark | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
/ v2 | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Toolathlon | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
AutomationBench | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
AA-Briefcase | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
GMMLU | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
MILU | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
INCLUDE | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
BioMysteryBench | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
LatchBio Bioinformatics | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Structural Biology/ open_ended | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
ProteinGym/ hard | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Organic Chemistry | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Protocol Troubleshooting | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported | |
SHADE-Arena | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
LinuxArena | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
ExploitBench | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
OSS-Fuzz | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Firefox 147 | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Long-form virology tasks/ task_2 | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Black-box RNA Sequence Design | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Black-box RNA Sequence Design | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Black-box RNA Sequence Design | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
Black-box RNA Sequence Design | other | mentioned | — | in-context iterationmissing: shot countmissing: languagemissing: training state | self-reported |
Black-box RNA Sequence Design | other | mentioned | — | in-context iterationmissing: shot countmissing: languagemissing: training state | self-reported |
AAV_Capsid_Packaging_Prediction | other | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |
| safety | mentioned | — | missing: shot countmissing: methodmissing: languagemissing: training state | self-reported |