← Models

Claude Sonnet 4.5

Anthropic · Claude 4

Previous balanced Claude Sonnet snapshot, now mainly useful for pinned deployments and historical comparison.

Part of Claude Sonnet family · Other versions: Claude Sonnet 5 , Claude Sonnet 4.6
Type
multimodal
Context
200K tokens
Max Output
64K tokens
Status
legacy
Input
$3/1M tok
Output
$15/1M tok
API Access
Yes
License
proprietary
reasoning coding multimodal tool-use balanced
Released September 2025 · Updated July 6, 2026

Overview

Freshness note: Model capabilities, limits, and pricing can change quickly. This profile is a point-in-time snapshot last verified on July 6, 2026.

Claude Sonnet 4.5 is Anthropic’s previous default production Sonnet model. It remains relevant for teams that pinned deployments before the Sonnet 4.6 launch or want a stable historical comparison against the current default.

Anthropic’s lifecycle page still lists claude-sonnet-4-5-20250929 as active with no tentative retirement sooner than September 29, 2026. Signal Lens marks it legacy because Sonnet 5 is the current Sonnet recommendation for new Claude builds.

Capabilities

The model is effective in structured reasoning, technical writing, implementation planning, and tool-enabled assistant workflows. It is a strong fit for mixed enterprise tasks where instruction fidelity and predictable behavior matter more than frontier benchmark chasing.

Technical Details

Anthropic’s Sonnet 4.5 launch positioned it as the best coding model in the world at the time, with strong agent, coding, and computer-use behavior. In July 2026, the more important context is comparative: Sonnet 4.5 is still available, but Sonnet 5 is now the current default and the model Anthropic recommends for most new Claude builds.

Current Anthropic versioning docs list the full Claude API model ID as claude-sonnet-4-5-20250929, with claude-sonnet-4-5 as a convenience alias. It keeps a 200K context window and 64K max output profile.

Pricing & Access

Current Anthropic pricing for Sonnet 4.5 remains 3per1Minputtokensand3 per 1M input tokens and 15 per 1M output tokens. Prompt caching is listed at 3.75per1Mtokensfor5−minutecachewrites,3.75 per 1M tokens for 5-minute cache writes, 6 for 1-hour cache writes, and 0.30forcachehitsandrefreshes;BatchAPIpricingis0.30 for cache hits and refreshes; Batch API pricing is 1.50 input and $7.50 output per 1M tokens. Access depends on whether your stack still exposes the pinned Sonnet 4.5 ID or alias, so teams should verify exact model availability before rollout. For new work, Sonnet 5 is usually the better default.

Best Use Cases

Good choice for engineering copilots, technical documentation workflows, analysis-heavy support tooling, and mid-to-high complexity product planning, especially when you need continuity with earlier Sonnet 4.5-based evals or workflows.

Comparisons

Compared with Claude Sonnet 5, Sonnet 4.5 is an older pinned route with a lower current upside. Compared with Claude Opus 4.8, it is more cost-efficient but less powerful on the hardest reasoning tasks. Compared with GPT-5.5, selection often comes down to ecosystem fit and workflow preference.