Claude Sonnet 5
Claude Sonnet 5 is a chat model from Anthropic, part of the Claude family's Sonnet line, measured at roughly 65 output tokens per second with about 0.8s to first token.
API Pricing
Cheapest on Anthropic — 38% below avg| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $1.00 | $5.00 | $0.200 | |
| $1.00 | $5.00 | $0.100 | |
| $2.00 | $10.00 | - | |
| $2.00 | $10.00 | $0.200 | |
| $2.00 | $10.00 | $0.200 |
Prices updated daily. Last check: Aug 20, 2026
Claude Sonnet 5 pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- GPQA Diamond80.0%
- Humanity's Last Exam19.0%
Coding
- SciCode48.6%
Agentic & Tool Use
- Terminal-Bench v2.175.3%
- τ-bench Banking15.7%
Instruction & Long Context
- Long-Context Reasoning66.0%
Benchmarks measured Aug 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- Anthropic
- Modalities
- Text
Capabilities
- Tool Calling
- No
- Open Source
- No
Strengths & Limitations
Strengths
- Measured output throughput of about 65.2 tokens per second in third-party testing, giving a predictable streaming rate for chat UIs
- Time to first token around 800 ms, which keeps interactive responses feeling responsive
- Part of Anthropic's Claude family, so it fits existing Anthropic-compatible client code and prompt patterns
- Sonnet-tier positioning targets a middle ground between Anthropic's smaller and larger model lines rather than optimizing for one extreme
- Available through inference APIs, so no GPU capacity needs to be provisioned to use it
- Listed by multiple providers where applicable, allowing per-token cost comparison in the table above
Limitations
- We do not have a confirmed context window on file for this model, so long-document limits should be verified with the provider
- No benchmark suite scores (reasoning, coding, math) are tracked in our database for this entry
- Modality support beyond text is not confirmed in our data — check the provider's model card before sending images or audio
- At roughly 65 output tokens/second, very long generations take meaningful wall-clock time compared with faster small models
- Availability and served configuration can differ between providers, so behavior may not be identical everywhere
Key Features
About Claude Sonnet 5
Common Use Cases
Claude Sonnet 5 fits general-purpose text workloads where a mid-tier model's balance of quality and serving cost matters: conversational assistants, drafting and editing prose, summarization, question answering over supplied text, and code explanation or review. Its sub-second time to first token suits user-facing chat and IDE-style assistants where the perceived delay before output starts matters more than raw generation rate; its mid-60s tokens-per-second throughput is adequate for typical response lengths but less well suited to jobs that emit tens of thousands of tokens per call under tight deadlines. For workloads that depend on a specific context length, structured output mode, or non-text input, confirm those capabilities against Anthropic's documentation first, since our database does not record them for this entry.
Frequently Asked Questions
How much does Claude Sonnet 5 cost?
Cost depends on which provider serves the model and how the pricing is structured — per input token, per output token, cached input, or batch discounts. Rates change frequently and differ between vendors, so check the pricing table on this page for current figures rather than relying on a quoted number.
What is Claude Sonnet 5 best used for?
General-purpose chat and text work: assistants, drafting and rewriting, summarization, Q&A over provided material, and code explanation. Its roughly 800 ms time to first token makes it a reasonable fit for interactive, streamed interfaces.
How fast is Claude Sonnet 5?
Third-party measurement from Artificial Analysis reports about 65.2 output tokens per second with a time to first token near 800 ms. Actual speed varies with provider, prompt size, and load.
What context window does Claude Sonnet 5 support?
We do not have a confirmed context window recorded for this model, so we do not publish one here. Check Anthropic's model documentation or your provider's model card for the supported limit.
How does Claude Sonnet 5 relate to other Claude models?
It sits in the Sonnet line, the name Anthropic uses for its middle model tier, between the smaller Haiku models and the larger Opus models. If you need maximum capability on hard reasoning tasks, compare it against an Opus-tier entry; if you need the lowest latency and cost for simple classification or routing, compare against a Haiku-tier entry.