Questions? hello@viberation.devGet supportBlogDocsChangelog
Get started

Anthropic's model family, strong at long-context coding work.

Claude is Anthropic's family of large language models. For vibe coders it is the model behind Claude Code and Claude.ai, and it handles large codebases and long instructions well.

At a glance — Claude Sonnet 4

Cost
$3.00 in · $15.00 out per 1M tokensMid-range
Memory
1M tokens about 1,500 pages
Reads
Text · Images · Files
Can do
  • Use tools (agents)
  • Structured output (not supported)
  • Thinking mode
Good at
Coding: better than 34%
Reliability
100.00% uptime, last 24h
Released
May 22, 2025
Advanced details
API model ID
anthropic/claude-sonnet-4
Max output
64K tokens
Cached input
$0.30 per 1M
Prompts over 200K
$6.00 in · $22.50 out per 1M
Knowledge cutoff
2025-01-31
Tokenizer
Claude
Coding index
37.6

Providers

The same model, hosted by different companies. OpenRouter picks one per request and falls back to the next if it fails.

ProviderInput /1MOutput /1MCached /1MMax outputUptime 24h
Amazon Bedrock · eu-west-1$3.00$15.00$0.3064K99.96%
Google · global$3.00$15.00$0.3064K100.00%
Google · europe$3.00$15.00$0.3064K
Amazon Bedrock$3.00$15.00$0.3064K99.99%
Google$3.00$15.00$0.3064K

Supported parameters

  • include_reasoning
  • max_tokens
  • reasoning
  • stop
  • temperature
  • tool_choice
  • tools
  • top_k
  • top_p

Live from OpenRouter, refreshed hourly. Scores from Artificial Analysis.

Official apps

Use it in

Pairs well with

Frameworks

Templates

Starter prompts

Copy one, fill in the brackets, and paste it into your tool.

  • Write a CLAUDE.md for this project · context

    Read this repository and write a CLAUDE.md at its root for future sessions. Keep it under 60 lines and include only what you could not work out quickly from the code: - How to install, run, test and build (exact commands) - Where things live, and any folder with a rule attached - Conventions a newcomer would get wrong - Things never to do here Ask me about anything you are unsure of instead of guessing.

  • Find the root cause, not a patch · debugging

    Something is broken. <symptom> [What you see, and the exact error message] </symptom> <steps> [How to make it happen] </steps> Do not fix anything yet. Find every place that calls the code involved, work out the actual cause, and explain it in two or three sentences. Then propose the smallest fix at the cause, plus one check that would have caught it.

  • UI that does not look AI-generated · design

    Design [the page or component] for [who it is for]. Commit to one clear visual direction and tell me what it is in one line before you build. Avoid the defaults every AI tool reaches for: purple gradients, one font everywhere, three identical feature cards, emoji as icons. <constraints> - Use the design tokens and components already in this project - Mobile first; nothing may scroll sideways at 375px wide - Text must meet WCAG AA contrast in light and dark mode </constraints>

  • Plan before any code · planning

    Before you write any code, read the files this touches and reply with a short plan: 1. What you will change, file by file 2. What could break, and why 3. How we will check it works Do not edit anything until I say go. <task> [Describe the feature or fix in a sentence or two] </task>

Key info

Pricing
Freemium
Category
Models
Best for
Beginner

Guided walkthroughs

Related reading

More in Models

  • OpenAI's model family, behind ChatGPT and Codex.

    Models
  • Google DeepMind's multimodal model family.

    Models