For production use through the Anthropic Claude API, Anthropic’s featured lineup is Claude Haiku 5.5, Sonnet 5.5, Opus 5.5, and Fable 5.1. Anthropic positions Haiku for high-volume, latency-sensitive work; Sonnet for a balance of speed and capability; Opus for long-running agentic coding and knowledge work; and Fable for demanding reasoning and long-horizon agentic work. These are starting points, not guarantees of performance on your workload. Anthropic recommends starting with Opus 5.5 for most workloads and considering Fable when task evaluations show Opus at higher effort is not enough. (Anthropic’s model overview)
Current featured Claude API models at a glance
These are Anthropic API model IDs, not Claude Code selectors or partner-provider identifiers. Anthropic lists a 1-million-token context window and a 128,000-token maximum output for all four models. Latency labels below are Anthropic’s relative comparisons, not response-time guarantees. (Anthropic’s model overview)
| Model | Anthropic API ID | Documented fit and relative latency | Context / max output | Base API price in USD per million tokens (input / output) |
|---|---|---|---|---|
| Claude Haiku 5.5 | claude-haiku-5-5 |
High-volume, latency-sensitive work; fastest | 1M / 128K | $0.10 / $0.50 for prompts up to 100,000 tokens; $0.50 / $2.50 for prompts over 100,000 tokens |
| Claude Sonnet 5.5 | claude-sonnet-5-5 |
Everyday coding, agent, and enterprise workloads; fast | 1M / 128K | $2 / $10 |
| Claude Opus 5.5 | claude-opus-5-5 |
Long-running agentic coding and knowledge work; moderate | 1M / 128K | $4 / $20 |
| Claude Fable 5.1 | claude-fable-5-1 |
Demanding reasoning and long-horizon agentic work; slower | 1M / 128K | $10 / $50 |
Prices are Anthropic’s listed base Claude API rates. Haiku’s rate depends on prompt length: its overview gives separate rates for prompts up to 100,000 tokens and prompts over that threshold. Check Anthropic’s pricing documentation for how the applicable tier is calculated for your request. (Anthropic pricing)
Which model should you choose?
- High-volume classification, extraction, or routing: Evaluate Haiku 5.5 first. Anthropic positions it for latency-sensitive, high-volume tasks and lists it as the fastest model. Check the prompt-length pricing tier for your traffic. (Haiku 5.5 overview)
- Everyday coding, agent tasks, and knowledge work: Evaluate Sonnet 5.5. Anthropic describes it as a combination of speed and intelligence for everyday workloads. (Sonnet 5.5 overview)
- Long-running agentic coding or knowledge work: Evaluate Opus 5.5. Anthropic positions it for these workloads and describes its relative latency as moderate. (Opus 5.5 overview)
- Demanding reasoning or long-horizon agentic work: Consider Fable 5.1 if your evaluations show Opus 5.5 at higher effort still does not meet your requirements. Anthropic describes Fable as slower and more expensive, so test whether its additional capability matters for your task. (Fable 5.1 overview)
For schema-constrained extraction, all four models in this comparison are listed as supported for Structured Outputs: claude-haiku-5-5, claude-sonnet-5-5, claude-opus-5-5, and claude-fable-5-1. Anthropic documents JSON outputs through the Messages API; use output_config.format to specify the JSON format. (Structured Outputs documentation)
Compare cost with a consistent workload
API cost depends on input and output tokens. For illustration, assume a request uses 100,000 input tokens and 10,000 output tokens, with base rates and no caching or other adjustments:
| Model | Illustrative token cost |
|---|---|
| Haiku 5.5 | $0.015 |
| Sonnet 5.5 | $0.30 |
| Opus 5.5 | $0.60 |
| Fable 5.1 | $1.50 |
Calculation: (input tokens ÷ 1,000,000 × input rate) + (output tokens ÷ 1,000,000 × output rate). The Haiku calculation uses the listed rates for prompts up to 100,000 tokens. These figures are arithmetic from Anthropic’s published rates—not a benchmark or a claim of equivalent quality. Actual cost depends on token usage and applicable pricing conditions. (Anthropic pricing)
For recurring prompts, Anthropic lists separate prompt-caching prices. For work that can run asynchronously, its Message Batches API offers a 50% discount on eligible input and output usage. Check current eligibility and terms before forecasting savings. (Anthropic pricing)
Production checks before you commit
- Test representative requests. Compare candidate models on your real task types and success criteria, such as accuracy, valid outputs, latency, and total token cost. A model’s positioning is not proof of its performance on your workload.
- Choose among the documented Structured Outputs candidates. Anthropic’s supported-model list includes all four IDs in this guide, and its documentation marks the feature available through the Messages API. Use
output_config.formatfor schema-constrained JSON and consult the documentation for schema requirements and limits. (Structured Outputs documentation) - Keep deployment channels distinct. The IDs in this article are for the Anthropic API. Partner platforms may use different identifiers and lifecycle schedules; check the relevant provider’s documentation for those deployments.
- Watch model lifecycle notices. Check Anthropic’s current deprecation documentation when maintaining an integration, and evaluate replacements against your application before migrating.
Bottom line
Choose by workload, then validate with your own evaluations: try Haiku for high-volume, latency-sensitive jobs; Sonnet for everyday work; Opus for long-running agentic tasks; and Fable for demanding work that still falls short on Opus at higher effort. All four models listed here are documented Structured Outputs candidates through the Messages API. Before rollout, test quality and latency on representative requests, and calculate cost using the correct input and output rates—including Haiku’s prompt-length tier.
Sources
Models overview — Claude Platform Docs
Claude Haiku 5.5 — Claude Platform Docs
Claude Sonnet 5.5 — Claude Platform Docs
Claude Opus 5.5 — Claude Platform Docs
Claude Fable 5.1 — Claude Platform Docs