Claude Sonnet 4.5

Claude Sonnet 4.5 is a billed Anthropic snapshot, not the current coding flagship. Rechecked 2026-08-17: the Claude models overview still lists claude-sonnet-4-5 at 200k context and $3 / $15 per 1M tokens. The same table now leads with later lines, including Claude Opus 4.6 and Claude Opus 5. Do not call 4.5 "the world's best coding model" in 2026 copy.

A 1M-context beta is mentioned on that page for some Claude 4+ models. It is not the default 200k row. Recheck the id you actually call.

Key Features

• Still on the API: claude-sonnet-4-5 remains a current snapshot id.
• 200k context on the price table. Do not assume 1M unless you enabled that beta path.
• Reliable mid-tier: Cheaper than current Opus 4.6 ($5 / $25) and far cheaper than Opus 5.
• Successor lineup: Newer agent/coding work should compare Opus 4.6, Opus 5, and Claude Fable 5.

Use Cases

• Pinned production that cannot leave claude-sonnet-4-5 yet.
• Cost-sensitive Claude traffic that does not need Opus 5.
• Comparisons against GPT-5.2 at a similar previous-frontier layer.

Limitation: launch SWE-bench and "best coding model" claims are stale marketing. We did not repeat them as current fact.

Pricing

From the 2026-08-17 overview table, per 1M tokens:

Getting Started

1. Read the models overview.
2. Call claude-sonnet-4-5 only if you intend to stay on this snapshot.
3. For new agents, compare Claude Opus 4.6 first.
4. Recheck 1M-context eligibility before you promise a million-token window.

Frequently Asked Questions

Is Sonnet 4.5 still the coding flagship?

No. It is a previous snapshot that is still billed.

Is the context 1M?

The price table prints 200k. 1M is a separate beta note.

Same price as Opus 4.5?

No. Sonnet 4.5 is $3 / $15. Opus 4.5 is $5 / $25.

Alternatives

• Claude Opus 4.6: Later Anthropic snapshot on the same overview.
• Claude Opus 5: Current high Claude line on this catalog.
• GPT-5.6 Sol: OpenAI frontier if you are leaving Claude.

Tips

1. Quote $3 / $15 from the overview, not a September 2025 launch blog.
2. Link forward to 4.6 / 5 whenever you say "current Claude".
3. Do not reuse "world's best coding model" without a fresh first-party line.

Conclusion

Sonnet 4.5 is still a cheap, billed Claude snapshot at $3 / $15 and 200k context. It is not the model Anthropic now leads with. Start at the overview, then decide whether to stay pinned or move to Opus 4.6.