Claude Sonnet 4
Anthropic · released May 22, 2025
Claude Sonnet 4 significantly enhances the capabilities of its predecessor, Sonnet 3.7, excelling in both coding and reasoning tasks with improved precision and controllability. Achieving state-of-the-art performance on SWE-bench (72.7%),...
Specification
- Context window
- 200K
- Max output
- 64K
- Knowledge cutoff
- Jan 31, 2025
- Parameters
- Undisclosed
- Licence
- Proprietary
- Serving providers
- 1
- Moderated
- Yes
- Uptime
- 100.0%
Intelligence
16.6
60th percentile
Coding
—
Coding Index
Agentic
—
Agentic Index
Output speed
52 t/s
Median across providers
Latency
1.08s
Time to first token
Cost per task
$1.35
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- MMLU-Pro83.7%
- GPQA Diamond68.3%
- τ²-bench (Telecom)52.3%
- IFBench45.4%
- LiveCodeBench44.9%
- AA-LCR (long context)44.0%
- AIME 202538.0%
- Terminal-Bench Hard27.3%
- Humanity's Last Exam4.3%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| MMLU-Pro | 83.7% |
| GPQA Diamond | 68.3% |
| τ²-bench (Telecom) | 52.3% |
| IFBench | 45.4% |
| LiveCodeBench | 44.9% |
| AA-LCR (long context) | 44.0% |
| AIME 2025 | 38.0% |
| Terminal-Bench Hard | 27.3% |
| Humanity's Last Exam | 4.3% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GPT-5 MiniOpenAI16.8
- o4 Mini HighOpenAI16.7
- o4 MiniOpenAI16.7
- Claude Sonnet 4Anthropic16.6
- Step 3.5 FlashStepFun16.6
- Gemini 2.5 ProGoogle16.1
- V3.2DeepSeek16.0
- Gemini 3.1 Flash LiteGoogle15.6
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GPT-5 Mini | 16.8 |
| o4 Mini High | 16.7 |
| o4 Mini | 16.7 |
| Claude Sonnet 4 | 16.6 |
| Step 3.5 Flash | 16.6 |
| Gemini 2.5 Pro | 16.1 |
| V3.2 | 16.0 |
| Gemini 3.1 Flash Lite | 15.6 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $3
- Output / 1M tokens
- $15
- Cached input / 1M
- $0.3
- Blended 3:1
- $6
One task, estimated
$1.35
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,158
- Win rate
- 56.5%
- Strongest at
- Data visualisation
- Tournaments
- 170
Elo by category
Dot position on a 1,075–1,175 scale · Elo has no meaningful zero
- 1,158
- 1,148
- 1,141
- 1,140
- 1,130
- 1,076
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| Data visualisation | 1158 | 56.5% |
| Websites | 1148 | 52.4% |
| Game development | 1141 | 54.8% |
| 3D scenes | 1140 | 57.8% |
| UI components | 1130 | 57.9% |
| SVG | 1076 | 51.1% |