Claude 4 Sonnet
Claude 4 Sonnet is a chat model from Anthropic in the Claude Sonnet line, positioned as the mid-tier option between Anthropic's smaller Haiku models and its larger Opus models.
API Pricing
| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $3.00 | $15.00 | $0.300 |
Prices updated daily. Last check: Sep 24, 2026
Claude 4 Sonnet pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- MMLU-Pro83.7%
- GPQA Diamond68.3%
- Humanity's Last Exam4.3%
Coding
- LiveCodeBench44.9%
Math
- AIME 202538.0%
- AIME40.7%
- MATH-50093.4%
Agentic & Tool Use
- Terminal-Bench Hard27.3%
- τ²-bench52.3%
Instruction & Long Context
- IFBench45.4%
- Long-Context Reasoning44.0%
Benchmarks measured Sep 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
- Sits in Anthropic's Sonnet tier, the family's mid-point between Haiku and Opus, giving a middle option without switching model vendors
- Available from Anthropic directly and through multiple resellers and aggregators, so buyers can compare rates in the pricing table on this page
- Part of the fourth-generation Claude family, so it shares the Claude API surface and message format used by other Claude models
- Drop-in substitution within the Claude family is straightforward, making tier escalation or downgrade a configuration change rather than a rewrite
- Widely integrated into third-party coding tools and agent frameworks that already target Anthropic's API
- Multi-provider availability creates routing options for redundancy and regional deployment
Limitations
- Weights are not distributed for self-hosting; access is API-only through Anthropic or its resellers
- We do not track verified benchmark scores for this entry, so capability comparisons must be made against provider-published numbers
- Our throughput and time-to-first-token measurements for this model are unpopulated, so latency should be tested per provider
- Context window and maximum output length are not recorded in our metadata and can be capped differently by individual resellers
- As a mid-tier model, it is not positioned for the hardest reasoning workloads that Anthropic aims its Opus tier at, nor for the lowest-cost high-volume jobs the Haiku tier targets
Key Features
About Claude 4 Sonnet
Common Use Cases
Claude 4 Sonnet suits general production chat and text workloads where a mid-tier model is the intended cost and quality point: coding assistance and code review, summarizing or extracting from long-form documents, drafting and editing content, internal knowledge assistants, and the routine steps of agent workflows that escalate only difficult subtasks to a larger model. It is a reasonable default when a team is already building on Anthropic's API and wants one model to cover most traffic. Workloads dominated by very high request volumes and simple classification are usually better matched to Anthropic's Haiku tier, while tasks where accuracy on hard multi-step problems dominates cost are typically evaluated against the Opus tier. Because our metadata does not confirm context limits or modality support for this entry, teams with long-document or image-input requirements should validate those against their chosen provider before committing.
Frequently Asked Questions
How much does Claude 4 Sonnet cost?
Pricing varies by provider and by pricing type — input versus output tokens, batch versus real-time, and any caching discounts a given vendor offers. Because Claude 4 Sonnet is resold by several platforms in addition to Anthropic, rates for the same model can differ. See the pricing table on this page for current per-provider figures.
What is Claude 4 Sonnet best used for?
It is best used as a general-purpose production chat model: coding help, document analysis and summarization, drafting and editing, customer-facing assistants, and the routine steps of agent workflows. It is the mid-tier choice in Anthropic's fourth-generation Claude family, so it is aimed at workloads that need more than a compact model but do not justify the largest tier.
How does Claude 4 Sonnet compare to Anthropic's Opus and Haiku models?
The three names denote tiers in the same family. Haiku is the compact tier Anthropic aims at high-volume, latency-sensitive work; Opus is the larger tier aimed at the hardest tasks; Sonnet sits between them. Switching between tiers is generally a model-ID change on the same API, so many teams benchmark two tiers on their own traffic and pick per route.
Can I run Claude 4 Sonnet on my own hardware?
No. Anthropic does not distribute Claude model weights, so access is through Anthropic's API or through cloud providers and aggregators that resell it. If self-hosting is a requirement, an open-weight model served on rented GPUs is the alternative path.
Does Claude 4 Sonnet accept image input or support a long context window?
Our database does not record a confirmed modality list or context window length for this entry, and a missing field here means unverified rather than unsupported. Check the documentation of the specific provider you plan to use, since resellers sometimes cap context or output length below the model's native limits.