Claude 3.5 Haiku: Complete Specifications, Pricing, API Access & Use Cases (2026)
What is Claude 3.5 Sonnet?
Anthropic announced Claude 3.5 Sonnet as the first release in the Claude 3.5 family and made it available through Claude.ai, the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI at launch.
Anthropic later released an upgraded Claude 3.5 Sonnet on October 22, 2024, emphasizing stronger coding, tool-use, and computer-use capabilities. The model is now best understood as a legacy reference model: Anthropic’s deprecation documentation lists both claude-3-5-sonnet-20240620 and claude-3-5-sonnet-20241022 as retired on October 28, 2025, with Claude Sonnet 4.6 recommended as the replacement.
Users still search for Claude 3.5 Sonnet because it was widely used for software engineering, document analysis, long-context prompting, image interpretation, and agent-style workflows before its retirement.
What Are Claude 3.5 Sonnet’s Key Specifications and Pricing?
| Field | Verified Value |
|---|---|
| Provider | Anthropic (as of June 2026) |
| Model Family | Claude 3.5 / Claude Sonnet (as of June 2026) |
| Model Type | Multimodal large language model (as of June 2026) |
| Release Date | June 21, 2024 official announcement date; upgraded version announced October 22, 2024 (as of June 2026) |
| Context Window | 200K tokens, based on Anthropic’s launch announcement (as of June 2026) |
| Input Pricing | $3.00 per 1M input tokens, historical launch/API pricing (as of June 2026) |
| Cached Input Pricing | Not confirmed from official sources as model-specific Claude 3.5 Sonnet pricing as of June 2026 |
| Output Pricing | $15.00 per 1M output tokens, historical launch/API pricing (as of June 2026) |
| Pricing Unit | USD per million tokens / MTok (as of June 2026) |
| Modality Support | Text input, image input, text output (as of June 2026) |
| Supported Input Types | Text and images; computer-use beta used screenshots and tools for the upgraded model (as of June 2026) |
| Supported Output Types | Text and code-as-text; native image, audio, or video output not confirmed (as of June 2026) |
| API Access | Retired on Anthropic API on October 28, 2025; historical access included Anthropic API, Amazon Bedrock, and Vertex AI (as of June 2026) |
| Model ID | claude-3-5-sonnet-20240620 and claude-3-5-sonnet-20241022, both retired (as of June 2026) |
| Availability | Retired on Claude API; Anthropic recommends claude-sonnet-4-6 as replacement (as of June 2026) |
| Knowledge Cutoff | Not confirmed from official sources as of June 2026 |
| Rate Limits | Not confirmed from official sources as model-specific limits as of June 2026 |
| Fine-tuning Support | Not confirmed from official sources as of June 2026 |
| Streaming Support | Supported by Claude API generally, but current model-specific use is not applicable after retirement as of June 2026 |
| Batch API Support | Not confirmed as active for Claude 3.5 Sonnet as of June 2026 |
| Tool / Function Calling | Tool use and computer-use beta were verified for the upgraded Claude 3.5 Sonnet during active availability (as of June 2026) |
| Structured Output / JSON Mode | Not confirmed from official sources as model-specific JSON mode as of June 2026 |
| License / Usage Restrictions | Proprietary Anthropic model governed by applicable Anthropic service terms during availability; standalone model license not specified as of June 2026 |
The key pricing point is historical rather than current: Anthropic’s launch announcement priced Claude 3.5 Sonnet at $3 per million input tokens and $15 per million output tokens with a 200K-token context window, while the current Anthropic deprecation page marks its API model IDs as retired.
What Can Claude 3.5 Sonnet Do That Makes It Useful in Production?
Long-context document and code analysis
The 200K-token context window made Claude 3.5 Sonnet useful for reviewing long documents, multi-file code snippets, support histories, and technical specifications. The limitation is that long context does not guarantee perfect recall, so critical outputs still require validation.
Software engineering assistance
Anthropic highlighted coding improvements, code editing, code translation, and multi-step software workflows. The upgraded October 2024 model improved coding and tool-use benchmarks, but production teams still needed tests, code review, and security checks before deployment.
Vision-based understanding
Claude 3.5 Sonnet supported image input and was positioned for interpreting charts, graphs, screenshots, and imperfect images. That made it relevant for document processing and visual QA workflows, while not replacing specialist OCR, compliance, or computer vision systems.
Agent-style computer use
The October 2024 upgrade introduced computer use in public beta, allowing developers to direct Claude to interact with interfaces through screenshots, cursor movement, clicks, and typing. Anthropic explicitly described this capability as experimental and error-prone at that stage.
Teams comparing Claude 3.5 Sonnet with other production models often also evaluate GPT-4o model specs and API access, Gemini 2.0 Flash multimodal model capabilities, and lower-cost alternatives such as GPT-4o mini pricing and use cases.
What Are Claude 3.5 Sonnet’s Supported Modalities?
| Modality | Supported? | Notes | Source Status |
|---|---|---|---|
| Text input | Yes | Standard prompt and conversation input during active API availability | Verified from Anthropic announcement/API context |
| Image input | Yes | Used for visual reasoning, charts, screenshots, and imperfect image text | Verified from Anthropic announcement |
| Audio input | Not confirmed | No official Claude 3.5 Sonnet audio-input support confirmed in reviewed sources | Not confirmed as of June 2026 |
| Video input | Not confirmed | No official Claude 3.5 Sonnet video-input support confirmed in reviewed sources | Not confirmed as of June 2026 |
| Text output | Yes | Primary output format | Verified from model behavior and API context |
| Native image output | Not confirmed | Claude 3.5 Sonnet was not documented as a native image-generation model | Not confirmed as of June 2026 |
| Audio output | Not confirmed | No official model-specific audio-output support confirmed | Not confirmed as of June 2026 |
| Computer interaction | Beta capability for upgraded model | Public beta for computer use through API tools; experimental and error-prone | Verified from Anthropic October 2024 announcement |
Where Does Claude 3.5 Sonnet Fall Short?
The most important limitation in 2026 is availability: Anthropic lists Claude 3.5 Sonnet API model IDs as retired, so new production integrations should not depend on them. Anthropic recommends Claude Sonnet 4.6 as a replacement in its deprecation table.
Claude 3.5 Sonnet also carried general AI limitations. It could hallucinate, omit relevant details in long contexts, misunderstand images, or produce plausible but incorrect code. This is a general AI limitation and is not model-specific unless stated by the provider.
For computer use, Anthropic described the October 2024 beta as imperfect and noted that scrolling, dragging, and zooming could present challenges. Teams using agentic workflows needed sandboxing, permissions controls, monitoring, and human review for sensitive actions.
Claude 3.5 Sonnet was not a native audio, video, or image-generation model based on the reviewed official sources. For dedicated reasoning or math-heavy workloads, teams may compare it with specialized reasoning models such as OpenAI o3 or OpenAI o4-mini, but the comparison should be task-specific rather than framed as an overall ranking.
What Is Claude 3.5 Sonnet Best Used For?
| Use Case | Why Claude 3.5 Sonnet May Fit | Important Limitation |
|---|---|---|
| Legacy application maintenance | Useful for understanding older integrations that used claude-3-5-sonnet-20240620 or claude-3-5-sonnet-20241022 | Retired API IDs should be migrated |
| Long document summarization | 200K context supported large prompts during active availability | Long context still required verification |
| Code review and migration | Anthropic emphasized coding, code translation, and multi-step development tasks | Code needed tests and human review |
| Visual document analysis | Image input supported chart, screenshot, and imperfect-image interpretation | Not a replacement for specialist OCR/compliance systems |
| Agent workflow research | Computer-use beta was historically important for GUI-agent experimentation | Experimental and error-prone in the reviewed release |
How Does Claude 3.5 Sonnet Compare to GPT-4o and Gemini 2.0 Flash?
| Comparison Area | Claude 3.5 Sonnet | GPT-4o | Gemini 2.0 Flash | Scenario Fit |
|---|---|---|---|---|
| Provider | Anthropic | OpenAI | Google / Google DeepMind | Choice depends on vendor ecosystem |
| Model status in 2026 | Legacy; Anthropic API IDs retired | Legacy but still referenced in many GPT-4o comparisons | Legacy Gemini 2.x model; later Gemini versions exist | Check current availability before integration |
| Context window | 200K tokens historically | Commonly referenced at 128K tokens for GPT-4o-era API use | Google described Gemini 2.0 Flash with a 1M-token context window | Long-context needs may favor newer active models |
| Modalities | Text and image input, text output | Text and vision API access at launch; broader GPT-4o multimodal roadmap | Multimodal input with text output in early 2.0 Flash documentation | Choose by required input/output modality |
| Coding and agent workflows | Strong historical fit; upgraded version added computer-use beta | Broad general multimodal and tool-use workflows | Low-latency multimodal and native tool-use positioning | Evaluate on task-specific tests |
| API status | Anthropic API IDs retired | Check current OpenAI model list before use | Check current Gemini model list before use | Avoid hard-coding retired IDs |
This comparison is intentionally limited to verified or broadly documented model characteristics. It does not declare a winner. For current deployments, the practical decision is usually between active successor models rather than Claude 3.5 Sonnet itself.
How To Access Claude 3.5 Sonnet?
As of June 2026, Claude 3.5 Sonnet should be treated as a retired Anthropic API model rather than an active integration target. Anthropic lists claude-3-5-sonnet-20240620 and claude-3-5-sonnet-20241022 as retired on October 28, 2025, and recommends claude-sonnet-4-6 as the replacement.
During its active availability, Claude 3.5 Sonnet was accessible through Claude.ai, the Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI. Anthropic’s current API overview states that the Claude API is available at https://api.anthropic.com, uses the Messages API at POST /v1/messages, and requires authentication headers such as x-api-key or Authorization, anthropic-version, and content-type.
This page does not include executable Python or curl examples for Claude 3.5 Sonnet because the verified model IDs are retired and examples using them would not be reliable for current production use.
FAQs
What was Claude 3.5 Sonnet’s context window?
Claude 3.5 Sonnet had a 200K-token context window according to Anthropic’s launch announcement. As of June 2026, this is a historical specification because Anthropic lists the Claude 3.5 Sonnet API model IDs as retired.
How much did Claude 3.5 Sonnet cost?
Anthropic’s launch announcement priced Claude 3.5 Sonnet at $3 per 1M input tokens and $15 per 1M output tokens. Current production pricing is not applicable to the retired Anthropic API IDs.
Can developers still access Claude 3.5 Sonnet through the Anthropic API?
Anthropic’s deprecation documentation lists claude-3-5-sonnet-20240620 and claude-3-5-sonnet-20241022 as retired on October 28, 2025. New integrations should use an active replacement model.
What was Claude 3.5 Sonnet commonly used for?
Claude 3.5 Sonnet was commonly used for coding assistance, document analysis, long-context reasoning, visual understanding, and early computer-use workflows. These uses now mainly matter for legacy migration and historical comparison.
