HEAD-TO-HEAD · FLAGSHIP vs MID-TIER · VERIFIED SEPTEMBER 5, 2026

Claude-Fable-5 vs Claude-Sonnet-5: Which Is Actually Cheaper?

Claude-Fable-5 and Claude-Sonnet-5 sit at opposite ends of Anthropic's current lineup. Fable 5 is the flagship, priced at $10 per million input tokens and $50 per million output. Sonnet 5 is the mid-tier workhorse at $2/$10. Both share a 1-million-token context window, but the 5x price gap makes workload routing critical.

Spec	Claude Fable 5	Claude Sonnet 5
Input / 1M tokens	$10.00	$2.00
Output / 1M tokens	$50.00	$10.00
Context window	1M tokens	1M tokens

The output price gap

Output tokens drive the cost difference. Claude-Fable-5 charges $50 per million output tokens versus Claude-Sonnet-5's $10—a 5x multiplier that compounds quickly on any output-heavy workload. Input pricing follows the same ratio: $10 per million for Fable 5, $2 per million for Sonnet 5. For drafting, code generation, or long-form content, that output premium is where the billing gap opens widest. If your use case generates more tokens than it consumes, Sonnet 5's price advantage becomes decisive unless Fable 5's quality gains materially reduce retries, errors, or review time.

Context window

Both models support a 1-million-token context window and up to 128,000 output tokens, so capacity is not a differentiator here. Teams choosing between them should focus on quality, task complexity, and total workflow cost rather than context limits. The identical window sizes mean you can route between the two models based purely on task difficulty without re-architecting prompts or splitting context. Hybrid strategies—using Fable 5 for orchestration and Sonnet 5 for high-volume execution—are practical because both models handle the same context scale.

Worked example

At 100 million input tokens and 30 million output tokens per month, Claude-Fable-5 costs $2,500 total: (100M × $10/M) + (30M × $50/M) = $1,000 input + $1,500 output. Claude-Sonnet-5 costs $500 total for the same volume: (100M × $2/M) + (30M × $10/M) = $200 input + $300 output. Sonnet 5 delivers a $2,000 monthly saving—an 80% reduction—on this realistic workload. The gap narrows slightly if Sonnet 5's newer tokenizer produces ~30% more tokens for equivalent text, but Sonnet 5 remains dramatically cheaper across most production volumes.

Prices from the LLM Price Watch daily tracker, September 5, 2026. Prices change; use the calculator with your own usage for an exact comparison, or see the full price table for every tracked model.

Frequently asked questions

Is Claude-Sonnet-5 pricing permanent or introductory?

Claude-Sonnet-5's $2 input / $10 output pricing is now permanent. Anthropic originally announced it as introductory pricing through August 31, 2026, with a planned increase to $3/$15 on September 1. However, the company canceled that increase, making the lower rate the standard price going forward. This decision significantly narrows the long-term cost gap with Claude-Fable-5 compared to earlier expectations.

When does Claude-Fable-5 justify its 5x price premium over Claude-Sonnet-5?

Claude-Fable-5 justifies its premium when task quality materially affects downstream costs—reduced retries, fewer errors, less human review, or better first-pass results on complex coding and reasoning problems. If a wrong answer is expensive or Sonnet 5 requires multiple attempts where Fable 5 succeeds once, the per-token gap shrinks in total workflow cost. For routine, high-volume work where Sonnet 5 performs adequately, Fable 5's premium is hard to justify.

Can I mix Claude-Fable-5 and Claude-Sonnet-5 in the same workflow?

Yes, and hybrid routing is one of the most effective cost strategies. Both models share a 1-million-token context window, so you can route tasks between them without re-architecting prompts or splitting context. A common pattern is using Fable 5 for orchestration, planning, or the hardest reasoning steps, then delegating high-volume execution to Sonnet 5 workers. This preserves premium reasoning where it matters while controlling cost on repetitive subtasks.

How much does prompt caching reduce costs on Claude-Fable-5 and Claude-Sonnet-5?

Prompt caching cuts input costs by up to 90% for repeated context. Cache hits cost one-tenth of the base input price—$1 per million tokens on Fable 5, $0.20 per million on Sonnet 5. Writing to the cache costs more than a standard input token, but if you're re-sending the same long context across requests, caching is the largest single cost lever available. The relative savings are identical across both models, so caching doesn't change which model is cheaper, but it does reduce absolute spend on either one.