Ranking AI Models
Back to ranking

Rank 2

Kimi K3

Review

A useful alternative when I want to compare another model's approach.

Best for: Second opinions and alternative implementation ideas.

Tradeoff: I have less usage evidence for it than my first-ranked model.

Evidence at a glance

Active days
9
Usage events
354
First observed
Jul 18, 2026
Last observed
Aug 4, 2026

Detailed evidence

Total processed tokens
15.8M
API-equivalent estimate
$10.63
Uncached input
987.8K
Output
218.9K
Cache reads
14.6M
Cache writes
0
Most used source
pi
Reasoning subset
0

Total processed tokens include uncached input, cache reads, cache writes, and output. Cache reads can dominate the total, so the token breakdown is shown separately.

Costs are analytical API-equivalent estimates reconstructed from local provider telemetry. They do not represent subscription charges, subscription quotas, invoices, or confirmed money paid.

Daily history

Daily aggregate model usage reconstructed by Memex
DateSourcesUsage eventsTotal processed tokensAPI-equivalent estimate
pi282,091,827$1.41
pi472,100,358$1.23
pi20770,724$0.56
pi12287,850$0.33
pi623,516,288$1.79
pi842,960,949$2.32
pi10177,520$0.24
pi431,786,495$1.14
pi482,142,356$1.60