Anthropic · explicit caching
Claude Sonnet 4.6
The best balance of speed and intelligence in the Claude line - a 1M context and the same 90% cache-read discount at a third of Opus input pricing.
Specifications
At a glance.
| Provider | Anthropic |
| Family | claude-sonnet |
| Released | 2026-02 |
| License | Proprietary |
| Context window | 1M tokens |
| Max output | 64K tokens |
| Modalities | text, image, pdf |
| Tool calling | Yes |
| Reasoning mode | Yes |
| Caching | explicit |
| Batch discount | 50% off |
Measured by Zumik
What reuse looks like here.
What you actually pay once caching works.
At a typical 55% prefix reuse, a million input tokens on Claude Sonnet 4.6 effectively costs $1.51 instead of $3.00 - blending to roughly $4.89 with a 25% output share. Background work drops a further 50% on the batch tier.
Estimate it for your workloadRoutes through these aliases:
Call it
Same OpenAI client, this model.
from openai import OpenAI
client = OpenAI(base_url="https://api.zumik.ai/v1", api_key="zk_live_...")
r = client.responses.create(
model="claude-sonnet-4-6", # or an alias like auto.balanced
input="Draft a fix for the failing test.",
)
print(r.usage.input_tokens_cached) # confirm reuseSimilar models
Other options for these workloads.
Frequently asked
Claude Sonnet 4.6, answered.
How much does Claude Sonnet 4.6 cost?
Claude Sonnet 4.6 is $3.00 per million input tokens and $15.00 per million output tokens through Zumik. Cache reads are $0.30 per million, a 90% discount on input.
What is Claude Sonnet 4.6's context window?
Claude Sonnet 4.6 supports a 1M-token context window with up to 64K output tokens.
Does Claude Sonnet 4.6 support prompt caching?
Yes. Anthropic uses Explicit cache_control breakpoints caching. In the Zumik corpus, Claude Sonnet 4.6 shows a median cache capture of 94% on agent workloads.
Which Zumik aliases route to Claude Sonnet 4.6?
Claude Sonnet 4.6 is a candidate for the auto.balanced, code.balanced aliases, selected when it wins under current routing policy.
Run Claude Sonnet 4.6 with reuse measured.
Point an OpenAI client at Zumik and see exactly how much of this model's input you are reusing.