CLAUDE.md · price reverts to £5
| Feature | Haiku 4.5 Fast & Cheap | Sonnet 4.6 Recommended | Opus 4.7 Most Capable |
|---|---|---|---|
| Input price (per 1M tokens) | $0.80 | $3.00 | $15.00 |
| Output price (per 1M tokens) | $4.00 | $15.00 | $75.00 |
| Context window | 200,000 tokens | 200,000 tokens | 200,000 tokens |
| Extended thinking | ✗ | ✓ | ✓ Deepest |
| Complex coding | ~ Basic tasks | ✓ Excellent | ✓ Best-in-class |
| Long document analysis | ~ Good | ✓ Very good | ✓ Excellent |
| Response speed | Fastest | Fast | Slower (more reasoning) |
| Multi-step reasoning | Limited | ✓ Strong | ✓ Strongest |
| Prompt caching | ✓ | ✓ | ✓ |
| Batch API | ✓ | ✓ | ✓ |
| Vision (image input) | ✓ | ✓ | ✓ |
| Claude Code support | ~ | Default model | ✓ Available |
| Best for | Classification, simple Q&A, data extraction | Most production use cases | Frontier research, hardest problems |
| Monthly Usage | Haiku 4.5 cost | Sonnet 4.6 cost | Opus 4.7 cost |
|---|---|---|---|
| 1M input + 200K output tokens | $0.80 + $0.80 = $1.60 | $3.00 + $3.00 = $6.00 | $15.00 + $15.00 = $30.00 |
| 10M input + 2M output tokens | $16 | $60 | $300 |
| 100M input + 20M output tokens | $160 | $600 | $3,000 |
| With prompt caching (80% cache hit) | ~$0.42/M eff. | ~$0.90/M eff. | ~$4.50/M eff. |
Output usage estimated at 20% of input tokens (typical for production). Use the Claude Cost Calculator for your exact workload.
The rule of thumb: start with Sonnet 4.6, and only upgrade to Opus 4.7 when you can measure a concrete quality improvement on your specific task. Opus is 5× more expensive — that premium is worth it when the task requires frontier-level reasoning (complex mathematics, nuanced legal/medical analysis, research-grade synthesis). For most software development, content, and business tasks, Sonnet 4.6 already performs at the top of its class. Use the cost calculator to model the ROI before committing to Opus for production.
| Model | API Model ID | Use in Claude Code |
|---|---|---|
| Claude Haiku 4.5 | claude-haiku-4-5-20251001 | --model claude-haiku-4-5-20251001 |
| Claude Sonnet 4.6 Default | claude-sonnet-4-6 | (default, no flag needed) |
| Claude Opus 4.7 | claude-opus-4-7 | --model claude-opus-4-7 |
What is the difference between Claude Sonnet and Opus?
Claude Opus 4.7 is Anthropic's most capable model with the deepest reasoning and best performance on complex tasks. Claude Sonnet 4.6 is nearly as capable on most tasks but costs 5× less ($3/M vs $15/M input). For most production use cases — including complex coding, document analysis, and agentic workflows — Sonnet 4.6 is the recommended default. Opus is worth its premium for frontier-difficulty problems where Sonnet visibly underperforms.
Is Claude Sonnet good enough, or should I use Opus?
For most workloads, Claude Sonnet 4.6 is good enough — it's Anthropic's recommended balance of capability and cost. Upgrade to Opus 4.7 only when you can measure a concrete quality improvement on your specific task. Opus 4.7 is worth the 5× premium for: complex multi-hop research, frontier math, and high-stakes analysis where accuracy is critical and cost is secondary.
When should I use Claude Haiku instead of Sonnet?
Use Claude Haiku 4.5 when you need high throughput, low latency, or low cost for simpler tasks. Haiku at $0.80/M input is ideal for: text classification, simple summarization, structured data extraction, real-time chatbots, and any pipeline where you're doing millions of simple calls. If your task requires nuanced reasoning or complex coding, Sonnet is worth the cost difference.
Do all Claude models have the same context window?
Yes — all three models in the Claude 4 family (Haiku 4.5, Sonnet 4.6, Opus 4.7) support a 200,000 token context window (~150,000 words). This is a significant advantage over GPT-4o (128K context). You don't need to use a more expensive model just to get a larger context window.
How much does Claude Opus cost vs Sonnet?
Claude Sonnet 4.6: $3.00/M input tokens, $15.00/M output tokens. Claude Opus 4.7: $15.00/M input, $75.00/M output — exactly 5× more expensive. With prompt caching (90% cache hit rate on repeated system prompts), effective input costs drop to ~$0.30/M for Sonnet and ~$1.50/M for Opus.