Claude Sonnet 5.5
Anthropic · released
The faster, cheaper companion to Opus 5.5 for well-scoped everyday coding, bug fixing and producing documents, slides and spreadsheets. At maximum effort it approaches Opus on several benchmarks; Anthropic says Opus stays ahead on open-ended work that needs sustained judgement.
Best for: Everyday coding · Fast iteration · Documents and slides
- Intelligence Index
- 56.0
- #2 of 38 · at Max effort
- Price per 1M tokens
- $2 in $10 out
- Blended $4 at 3:1
- Context window
- 1M
- Up to 128K output
- Output speed
- 129 tokens/s
- $5.46 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall90 · #4 of 38
Based on 3 of 3 benchmarks
- Coding96 · #2 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work89 · #4 of 40
Based on 6 of 6 benchmarks
- Reasoning & science93 · #4 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability86 · #6 of 39
Based on 2 of 2 benchmarks
- Vision68 · #16 of 31
Based on 1 of 2 benchmarks
Reasoning effort trade-off
The same model at each effort setting. Lower settings give up a few points for a fraction of the cost and wait.
| Setting | Intelligence Index | Cost / task | First answer |
|---|---|---|---|
| Max | 56.0 | $5.46 | 7.7 min |
| xHighSweet spot | 51.9 | $2.01 | 36s |
| High | 46.8 | $0.88 | 13s |
| Medium | 40.8 | $0.48 | 2.1s |
| Low | 35.9 | $0.35 | 1.1s |
Measured by Artificial Analysis. First answer is the median wait until the first answer token, including thinking time. “Sweet spot” is the cheapest setting that keeps at least 90% of the top score for no more than 60% of its cost.
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#2 of 38· Max effort· Artificial Analysis
- Vals Index67.0%
#2 of 26· Max effort· $21.3 per test· Vals AI
- Arena · Text1471 ±10
#15 of 33· xHigh effort· 3,145 votes· Arena
Coding
- Terminal-Bench 4.063.6%
#1 of 38· Max effort· Artificial Analysis
- SciCode61.0%
#4 of 38· Max effort· Artificial Analysis
- Arena · WebDev1773 ±15
#3 of 34· xHigh effort· 2,150 votes· Arena
- Vibe Code Bench92.4%
#1 of 26· Max effort· Vals AI
- Code Migration70.1%
#1 of 26· Max effort· Vals AI
Agents & knowledge work
- GDPval-AA67.0
#2 of 39· Max effort· Artificial Analysis
- Arena · Agent+12.5 ±3.1
#3 of 29· Max effort· Arena
- Finance Agent v258.1%
#6 of 26· Max effort· Vals AI
#3 of 26· Max effort· Vals AI
- Legal Research Bench48.1%
#6 of 26· Max effort· Vals AI
- Tax Agent Bench73.4%
#3 of 26· Max effort· Vals AI
Reasoning & science
- Humanity's Last Exam55.0%
#4 of 39· Max effort· Artificial Analysis
- CritPt31.4%
#4 of 39· Max effort· Artificial Analysis
- ARC-AGI-2saturatingNot tested
- ARC-AGI-3Not tested
- GPQA DiamondsaturatingNot tested
Knowledge & reliability
#6 of 39· Max effort· Artificial Analysis
- AA-LCR82.7%
#13 of 39· Max effort· Artificial Analysis
Vision
- MMMU-ProNot tested
- Arena · Vision1268 ±15
#9 of 19· xHigh effort· 1,608 votes· Arena
Reported by Anthropic
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- Terminal-Bench 4.0 · Max effort70.6%
- SWE-bench Pro · Max effort81.3%
- OSWorld 2.1 · partial score80.1%
- CursorBench 4.0 · run by Cursor55.5%
Specifications
- Lab
- Anthropic
- Released
- Sep 28, 2026
- Availability
- Generally available
- Context window
- 1,000,000 tokens
- Max output
- 128,000 tokens
- Accepts
- Text, Images, PDFs
- Knowledge cutoff
- Jun 2026
- Weights
- Closed
- Reasoning control
- On by default · effort low to max
- API model ID
claude-sonnet-5-5
Pricing
- Input / 1M tokens
- $2
- Cached input / 1M
- $0.10
- Output / 1M tokens
- $10
- Blended (3:1)
- $4
- Measured cost / task
- $5.46
Full 1M context at standard rates. Batch API is half price.
Worth knowing
- Safeguards hand some cybersecurity and life-science requests to an older Opus model. Independent evaluators score what the product returns, so their results include those fallbacks.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.