AI

Fable 5.1 finishes more tasks and is often cheaper, but it does not always do better work

The new Fable model wins most comparisons with its predecessor and comes out cheaper in many tests. However, some testers note that the older Fable 5 often did better on ordinary projects and that the new model hallucinates more.

Bartosz Bielawski

Bartosz Bielawski

· 2 min read
Share
Fable 5.1 finishes more tasks and is often cheaper, but it does not always do better work

Signal in the data

Independent sources on this topic: 3 in the last week, 7 in 30 days, 8 in 90 days. We have tracked it since 2026-07-08; 58 claims in our database in total.

wk 260
wk 276
wk 280
wk 294
wk 300
wk 311
wk 320
wk 331
wk 340
wk 353
wk 364
wk 371

Number of independent sources per week (last 12 weeks).

What is happening

Anthropic has released Fable 5.1, the successor to Fable 5. Within two weeks, a series of tests had compared it with its predecessor and with GPT-6 Astra.

Two questions get the most attention: how much the work costs, and whether the new model really does better work.

Where the sources agree

Most tests favour the new version. In a head-to-head comparison, Fable 5.1 won 30 of 50 tests, while Fable 5 won only 17 (The AI Advantage).

Another recurring point is that 5.1 writes better, with more facts and less fluff. It is also better at front-end work and at using add-on capabilities known as skills.

Where they differ

Cost is where they differ most. Some report that work is a quarter cheaper, and nearly half the cost for agentic tasks (Nate B Jones). Others point out that the model generates more text and, in independent tests, costs about 20% more per task (AI LABS).

Quality ratings differ too. On one tester's own projects, the older Fable 5 did better on most tasks. Fable 5.1 won only the hardest one, though it also finished that one faster and more cheaply. The same tests found the new model more accurate but more prone to hallucination.

What to do about it

Do not assume the new version is better at everything. Compare both models on a few of your typical tasks and check the real cost.

If your prompts include rules that limit bold text, headings and lists, review them. Fable 5.1 uses less formatting on its own, so old rules may get in its way (Search Engine Journal).

Our take

Fable 5.1 is better at finishing hard tasks and at writing, and it is cheaper in many scenarios. However, the advantage is not universal, and because it is more prone to hallucination, you still need to check its output as you did before.

Confidence: medium - there are many tests, but they reach conflicting conclusions on cost and on quality for ordinary tasks. We will change our view if: A larger independent comparison shows Fable 5.1 beating its predecessor on typical tasks as well, or finds its cost per task to be clearly higher.

Sources

Read more