Today's price · per 1M tokens
Input
$2.00
per 1M tokens
Output
$10.00
per 1M tokens
Blended
$4.00
blended $/1M — 3:1 weighted input:output
Cached input
$0.100
5% of input — prompt caching
Price receipt
We read Anthropic's own pricing page on and found Claude Sonnet 5.5 listed with a price on the same row — that read is where the number above comes from, recorded as a price drop. We keep the page text we read; its content hash is 021bc760d5.
Overview
Anthropic's newest Sonnet and the second model of the Claude 5.5 family, launched 2026-09-28 as a faster, lower-cost complement to Opus 5.5. Anthropic says it is a clear upgrade over Sonnet 5, generates 30%+ faster and costs up to 30% less per task because it needs fewer tokens. List price is unchanged from Sonnet 5: $2/$10 per 1M tokens, cache reads $0.10 (5% of input, cut from $0.20 by 2026-10-07), cache writes $2.50; Batch API halves it to $1/$5; US-only inference is 1.1x. 1M-token context, 128K max output.
Capabilities
struck through = not supportedBenchmark performance
accuracy % · higher is betterSources: Vals AI Terminal-Bench 2.1 leaderboard (high effort; 10/267 tasks fell back to Sonnet 5 server-side) (Terminal-Bench 2.1) · Artificial Analysis (text-only, no tools, max effort, default fallback) (Humanity's Last Exam)
- AA Intelligence Index: 54.5 (max effort) (Artificial Analysis composite across reasoning, knowledge and coding evals)
- AA-Omniscience Index: 30.5 (max effort) (−100–100)
- GDPval-AA v2: 1837.8 ELO (Real-world work tasks, human baseline = 1000)
- Arena (arena.ai): 1472.9 ELO (Human preference)
- Arena WebDev: 1772.3 ELO (Human preference on building web apps)
- Arena Vision: 1262.7 ELO (Human preference on image-input prompts)
- Arena Agent: 0.16 (Score from real agent sessions (IPS); higher is better)
- Arena Agent cost per task: $3.35 (Median measured cost of one real Agent Arena task)
- Epoch Capabilities Index: 166.6 (CI 163.8–169.6) (Epoch AI's index fitted across many benchmarks)
- FrontierMath Tiers 1–3: 88.8% (max effort) (Research-level maths (funded by OpenAI, run by Epoch); best run across the effort settings Epoch tested)
- FrontierMath Tier 4: 80.5% (max effort) (Hardest research maths; best run across the effort settings Epoch tested)
Source: Artificial Analysis (artificialanalysis.ai)
- Humanity's Last Exam: 55% (max effort)
- AA-LCR (long-context reasoning): 82.7% (max effort)
- CritPt (research-level physics): 31.4% (max effort)
- Terminal-Bench 4.0: 63.6% (max effort)
- SciCode: 61% (max effort)
- Cost to run the AA Intelligence Index: $10.02 per task, 34 min per task (max effort)
- Measured by AA: 140.1 tokens/sec output, 246.27s to first token (max effort)
- Ranks #49 of 100 benchmarked models by average score
- Ranks #15 of 189 comparably-measured models by percentile score, across 6 independent measurements
- Ranks #131 of 189 comparably-tested models by normalized performance per dollar
- Strongest at Humanity's Last Exam — 55%, #10 of 81
140.1 tokens/sec output 246.27s latency to first token (TTFT)
Throughput: Artificial Analysis, measured 2026-11-14 · Source: Artificial Analysis (artificialanalysis.ai)
Source: Vellum LLM Leaderboard · updated · See full rankings →
Specifications
- Provider
- Anthropic
- Context window
- 1M tokens
- Modality
- text, image
- Parameters
- Proprietary
- Open source
- No — proprietary
- Released
- Status
- Current
- Last updated
- Tags
Availability verified: — listed on Anthropic's own page