What dropped
Anthropic released the Claude 4 model family in mid-2026. The lineup:
| Model | ID | Best for |
|---|---|---|
| Claude Opus 4.8 | claude-opus-4-8 | Complex reasoning, long context tasks |
| Claude Sonnet 4.6 | claude-sonnet-4-6 | Balanced speed + quality, daily use |
| Claude Haiku 4.5 | claude-haiku-4-5 | Fast, cheap, high-volume tasks |
What's new vs Claude 3
- Extended thinking on Opus — the model reasons step by step before answering
- Faster output on Sonnet — noticeable improvement in time-to-first-token
- Better instruction following across the board — fewer hallucinated refusals
- 200k context maintained on Opus and Sonnet
When to use which
Opus 4.8 — complex multi-step reasoning, document analysis, code review across large codebases. Use when quality matters more than cost.
Sonnet 4.6 — production apps, chat, coding assistants, summarization. Best balance.
Haiku 4.5 — classification, tagging, extraction at scale. Cheapest, fastest.
API quick start
import anthropic
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
messages=[{"role": "user", "content": "Explain Kafka consumer groups."}],
)
print(message.content[0].text)
Bottom line
If you're on Claude 3 Sonnet, upgrade to Claude Sonnet 4.6 — it's a straight improvement with no breaking API changes.