On June 9, 2026, Anthropic launched Claude Fable 5 — the most powerful model it had ever made publicly available. Days later, it pulled it offline following a U.S. government export directive. If your product depended on that one model, your product was down, through no fault of your own.
And even while it was up, Fable came with strings: sensitive queries (cybersecurity, chemistry, biology) were quietly answered by the older Opus 4.8 instead, and after the first two weeks it moved to usage-credit billing. Powerful, yes — but restricted, re-priced, and ultimately revocable by forces entirely outside your control.
The lesson isn't "avoid Fable." It's "don't bet on any single model."
This is the same story as every model deprecation, just faster and more dramatic. A model you pin your product to can be deprecated on a schedule, rate-limited, re-priced, safety-gated, or — as Fable just demonstrated — pulled offline overnight by a government directive. The fix isn't picking a different single model. It's not depending on one at all.
A Nest is your own model, made of many
On Antbase, a Nest is a named selection of LLMs that becomes its own virtual model. You pick the models — across providers — and ANT routes intelligently within that pool, with automatic fallback. It's exposed as a single OpenAI-compatible handle: nest:your-name. Your app calls one model string; behind it sits a pool you control.
Build a frontier Nest from the strongest models available — Anthropic's best, plus Gemini, Qwen3-235B, DeepSeek R1 — and call it as one model:
from openai import OpenAI
client = OpenAI(base_url="https://antbase.ai/v1", api_key="ant-...")
# nest:frontier = a pool you defined: Claude Opus 4.8, Gemini, Qwen3-235B,
# DeepSeek R1 ... ANT routes within it and falls back automatically.
resp = client.chat.completions.create(
model="nest:frontier",
messages=[{"role": "user", "content": "Hello"}],
)Now picture the Fable scenario again. One model in your Nest gets pulled offline. ANT routes the request to the next-best model in the pool. No 4xx storm, no incident, no code change — your product keeps answering. That's the entire point: the resilience lives in the Nest, not in any one model's continued existence.
So is a Nest "the alternative" to Fable?
Honestly: a Nest isn't a clone of one model's peak capability — it's a different, better bet. A single flagship gives you one model's ceiling and all of its risk. A Nest gives you the best available model for each request and survives any one of them disappearing. And nothing stops you from putting Anthropic's most powerful available model inside your Nest, alongside others — you get its strength when it's up, and the rest of the pool when it isn't. The difference is who decides what your app runs on. With a Nest, that's you — not a vendor's roadmap, and not a government directive.
Create one in the dashboard (Models → Create Nest), pick your models, and it's instantly live as nest:your-name. Or compare the frontier models you'd put in it, side by side, in the playground:
Models will keep launching, getting restricted, and disappearing — sometimes within a single week. Build on a pool you control, and none of it is your emergency.
