Breaking

Claude Fable 5.1 Takes 18 Minutes and $12.60 Versus Kimi K3's 10 Minutes and $6.95 on the Same Max-Reasoning Prompt

September 3, 2026 at 12:52 EDT

Foundation Models
Open Source

A same-prompt, max-reasoning run of Anthropic's newly released Claude Fable 5.1 finished in 18 minutes at $12.60, while Moonshot AI's Kimi K3 completed the identical task in 10 minutes at $6.95. Both answers were judged impressive, leaving a roughly 1.8-times gap in time and spend as the result that is hardest to ignore.

Sep 3, 2026 · Frontier models

Same prompt. Different bill.

On a shared max-reasoning task, Claude Fable 5.1 took 18 minutes and $12.60. Kimi K3 finished in 10 minutes for $6.95. Both outputs were judged impressive.

Max-reasoning run, drawn to scale

Each stack uses 9 blocks versus 5 — about 1.8×, matching both the time gap and the dollar gap.

Wall clock

18 min

Claude Fable 5.1

10 min

Kimi K3

Run cost

$12.60

Claude Fable 5.1

$6.95

Kimi K3

Takeaway: On this prompt, Fable 5.1 cost and waited about 1.8× as much as Kimi K3.

$10 / $50

Fable 5.1 list input / output per 1M tokens

$3 / $15

Kimi K3 list input / output per 1M tokens

$0.25

Fable 5.1 cache read per 1M tokens

3.3×

Fable 5.1 list-price premium on uncached tokens

What landed

Both outputs looked strong. Developers say a near two-times cut in time and spend for closely matched reasoning is the part that matters.

What it is not

One prompt is not a leaderboard. Fable 5.1 still leads several hard tests, and max-effort output is still billed at $50 per million tokens.

List prices still dwarf this one-run gap. Cache is the exception: Fable 5.1 now undercuts Kimi K3 on rereads.

Fable 5.1 cache

$0.25 / 1M

Kimi K3 cache

$0.30 / 1M

Kimi K3 scale

2.8T MoE, open weights

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.