Home → Guides → Claude Fable 5 pricing: Fable 5 vs 5.1, cost per token and the real difference
Same headline price, very different bills. The two Fable models differ in one line of the rate card — cache reads — and in an agent session that line is most of what you pay.
Per million tokens, from Anthropic's published rates:
| Model | Input | Output | Cache read | Cache write (5 min) |
|---|---|---|---|---|
claude-fable-5 | $10.00 | $50.00 | $1.00 | $12.50 |
claude-fable-5-1 | $10.00 | $50.00 | $0.25 | $12.50 |
Fable 5 charges the usual Claude cache-read rate, a tenth of input. Fable 5.1 charges a fortieth (0.025×) — the lowest cache-read multiplier in the Claude line. Everything else is identical.
On this platform the same models cost:
| Model | Input / 1M | Output / 1M | Share of list price |
|---|---|---|---|
claude-fable-5 | $4.50 | $22.50 | 45% |
claude-fable-5-1 | $4.50 | $22.50 | 45% |
Across all 14 billed claude-fable-5 calls that have gone through this gateway, priced at list:
Total $2.518, and it matches what the system billed. Fresh input and output together are about 6% of it. The rest is cache traffic: the conversation and tool definitions being written to the cache and read back on every turn. That is normal for Claude Code and other agents, and it is why the headline $10 / $50 tells you very little.
Re-price those 840,446 cache-read tokens at 5.1's $0.25 and they cost $0.210 instead of $0.840. The session drops from $2.518 to $1.888 — 25% cheaper for an identical workload, with no change to your code beyond the model name.
A simpler per-turn example: an agent turn that reads a 100,000-token cached prefix, adds 5,000 fresh input tokens and writes 2,000 output tokens.
claude-fable-5: $0.100 cache read + $0.050 input + $0.100 output = $0.250claude-fable-5-1: $0.025 cache read + $0.050 input + $0.100 output = $0.175Thirty percent less per turn. The longer the cached prefix, the bigger the gap; for a short one-off prompt with no cache there is no difference at all.
The model strings use hyphens: claude-fable-5 and claude-fable-5-1. Writing claude-fable-5.1 with a dot returns a model-not-found error.
For Claude Code, set the base URL to https://apiclan.us (no /v1) and choose the model:
export ANTHROPIC_BASE_URL=https://apiclan.us
export ANTHROPIC_AUTH_TOKEN=sk-your-key
claude --model claude-fable-5-1With the Anthropic SDK or a raw request, post to https://apiclan.us/v1/messages with your key in the Authorization: Bearer header.
tool_choice of any or a named tool is rejected. Keep auto and say in the prompt when the tool applies.temperature, top_p or top_k are rejected — and many frameworks send a default temperature you never set.Code written for older Claude models usually hits at least one of these. The fixes are covered in this troubleshooting note.
If your workload re-reads a large context — coding agents, long chats, document Q&A — Fable 5.1 is the cheaper of the two at the same list price, and nothing else changes. Fable 5 only makes sense where you have tested a specific behaviour on it and want to keep it. A deeper breakdown of 5.1's cache rate against a real invoice is in the Fable 5.1 model guide.
No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.
Read the 30-second quickstartLanguage: English · 简体中文 · 繁體中文 · 日本語 · 한국어 · Deutsch · Français · Español · Türkçe · Português · Italiano · Bahasa Indonesia · Tiếng Việt · Bahasa Melayu · Polski · Čeština · فارسی · Русский
Prices on this page are filled in automatically from live billing data. Third-party terms are quoted from that party's own documentation.