The fable-5 model — what's new, what's improved, and why it's a steal through LiteAI

The fable-5 model is the new flagship in Anthropic's lineup, available through LiteAI at a flat $10 per 1M tokens. Anthropic's direct price is $5 per 1M input and $25 per 1M output. The gap is real on output — and in this article we break down what you're paying for and how much you save.

What fable-5 is

fable-5 is the next-generation Anthropic model, succeeding previous flagships in tasks that demand reasoning, code work, and long-context understanding. It's available through LiteAI as a dedicated plan: 10M tokens for $100, fable-5 only.

What's improved

Reasoning and instruction following

fable-5 is noticeably better at following multi-step instructions. Where the previous model might skip a step in a long prompt or mix up the order — fable-5 holds the structure. This matters for agentic tasks: when you give an agent a 10-step plan, it executes all 10, not 7.

Code quality

Programming is one of Anthropic's core strengths, and fable-5 raises the bar:

• Fewer hallucinated API calls: the model less often invents non-existent methods or library parameters.
• Better large-file refactoring: fable-5 retains the context of an entire file and makes changes coherently, without breaking adjacent functions.
• Faster style adoption: if your codebase has conventions (naming, indentation, patterns), the model picks them up quicker.

Long context

fable-5 handles long context more effectively. In practice: if you feed 50 pages of documentation in a single request, the model doesn't "forget" the beginning by the time it reaches the end. Previously, quality degraded with large contexts — now the degradation is minimal.

Tool use and function calling

fable-5 calls external tools more reliably: fewer JSON schema errors, correct handling of call results, better recovery from failures. For agents that query databases, APIs, and file systems — this means fewer stalls and fewer manual fixes.

Fewer false refusals

Previous models sometimes refused to answer legitimate queries, mistaking them for unsafe ones. fable-5 is better at telling real risks apart from false positives — fewer "I can't help with that" responses on perfectly normal questions.

Price: Anthropic direct vs LiteAI

Anthropic's direct prices for fable-5:

	Anthropic direct	LiteAI
Input (1M tokens)	$5	$10
Output (1M tokens)	$25	$10
Blended (input + output 1:1)	$15	$10

On output — 2.5× cheaper. Blended — 1.5× cheaper. Input is priced at parity, but LiteAI gives you one flat rate for input and output: you pay the same regardless of direction, no need to track token types.

This is not a typo. LiteAI offers fable-5 at a flat $10 per 1M tokens — regardless of whether it's input or output. With Anthropic, output costs 5× more than input, and output is where most of your spend goes in the majority of workloads.

Example calculation

Take a typical scenario: an agent that processes documents and generates reports. 3M input tokens + 2M output tokens per day.

	Anthropic direct	LiteAI
Input: 3M × $5	$15	3M × $10 = $30
Output: 2M × $25	$50	2M × $10 = $20
Daily total	$65	$50
Monthly	~$1,950	~$1,500

Savings — 1.3×, or ~$450 per month. The more output-heavy your workload is, the more you save.

Why so cheap

LiteAI isn't reselling individual API calls. We buy tokens in bulk from Anthropic and sell them through our own proxy (Bifrost). You get a single API key sk-bf-… that works with an OpenAI-compatible endpoint. No difference in quality — same fable-5, same answers, same speed.

The only difference is price and payment method: USDT via Bitbanker — no foreign cards, no KYC, no VPN.

How to connect fable-5

Via API

import openai

client = openai.OpenAI(
    api_key="sk-bf-...",
    base_url="https://api.liteai.tech/v1"
)

response = client.chat.completions.create(
    model="fable-5",
    messages=[{"role": "user", "content": "Explain how RAG works"}],
    max_tokens=2048,
    stream=True,
)

The same code works with curl, node-fetch, any SDK — everything compatible with the OpenAI API is compatible with LiteAI.

Via Claude Code

export ANTHROPIC_API_KEY="sk-bf-..."
export ANTHROPIC_BASE_URL="https://api.liteai.tech/anthropic"
claude

fable-5 will appear in the model list. Select it — and Claude Code will use fable-5 for all tasks.

Who fable-5 is for

• Dev teams — refactoring, code review, writing tests. fable-5 understands the whole project and makes precise changes.
• Agentic tasks — multi-step plans, tool use, external API calls. Fewer stalls, more autonomy.
• Large document analysis — contracts, specs, technical docs. Long context without quality loss.
• Production workloads with high output spend — if output tokens are your main cost, savings on fable-5 through LiteAI are maximal.

Who it's not for

If you need multimodality (voice, images, realtime) — fable-5 doesn't do that, go with OpenAI. If you need the cheapest option for mass simple tasks — DeepSeek V3 is cheaper. fable-5 is when you want flagship quality at a fair flat price with no VPN and no foreign cards.

Bottom line

fable-5 is a tangible step forward in reasoning quality, code, and long-context handling. Through LiteAI you get that step at $10 per 1M tokens — 1.5× cheaper than Anthropic's blended direct price, and 2.5× cheaper on output.

The 10M token package costs $100. That's enough for a week of intensive development or a month of a production agent. Payment is USDT, and the key arrives in 30 seconds.