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
$0.200
10% of input — prompt caching
How this price is scoped: Standard pricing, per 1M tokens — not a promotional rate. $2/$10 was announced at launch as introductory pricing through Aug 31 2026, but Anthropic's pricing page now states that it "is now the standard price" and that "the previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur". There is therefore no reversion date to watch. Verified 2026-08-13 against platform.claude.com/docs/en/about-claude/pricing.
Price receipt
We read Anthropic's own pricing page on and found Claude Sonnet 5 listed with a price on the same row — that read is where the number above comes from. We keep the page text we read; its content hash is 89a12b8bd5.
Overview
Anthropic's Sonnet from 2026-06-30, still sold but succeeded by Sonnet 5.5 on 2026-09-28. Its most agentic Sonnet at launch — performance approaching Opus 4.8 at a fraction of the price, strong on reasoning, tool use, coding and autonomous agent workloads. $2/$10 per 1M tokens — first announced as launch pricing, now Anthropic's standard price; the increase once scheduled for Sep 1 2026 will not happen.
Capabilities
struck through = not supportedBenchmark performance
accuracy % · higher is betterSources: Vellum LLM Leaderboard (GPQA Diamond, SWE-Bench Verified, Terminal-Bench 2.1, BrowseComp, OSWorld-Verified, Humanity's Last Exam, AutoBench)
- GDPval-AA v2: 1469.8 ELO (Real-world work tasks, human baseline = 1000)
- Arena (arena.ai): 1460.8 ELO (Human preference)
- Arena WebDev: 1539 ELO (Human preference on building web apps)
- Arena Vision: 1260.8 ELO (Human preference on image-input prompts)
- Arena Agent: 0.02 (Score from real agent sessions (IPS); higher is better)
- Arena Agent cost per task: $0.95 (Median measured cost of one real Agent Arena task)
- Epoch Capabilities Index: 156.9 (CI 154.8–159.2) (Epoch AI's index fitted across many benchmarks)
- FrontierMath Tiers 1–3: 65.6% (max effort) (Research-level maths (funded by OpenAI, run by Epoch); best run across the effort settings Epoch tested)
- FrontierMath Tier 4: 29.3% (max effort) (Hardest research maths; best run across the effort settings Epoch tested)
Source: Artificial Analysis (artificialanalysis.ai)
- Ranks #11 of 100 benchmarked models by average score
- Ranks #25 of 189 comparably-measured models by percentile score, across 8 independent measurements
- Ranks #140 of 189 comparably-tested models by normalized performance per dollar
- Strongest at GPQA Diamond — 96.2%, #1 of 71
56.3 tokens/sec output 20.69s latency to first token (TTFT)
Throughput: Artificial Analysis via Vellum, recorded 2026-07-25 · 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