kimi-k3
Ranked 2 of 14 overall. Strongest on discovery & prediction at 98.9, weakest on credit analysis at 83.5.
Score
90.1
rank 2 of 14
Latency
400s
median per task
Cost
$0.332
per task at list prices
Capability profile
Mean score per category on question-answering, 0–100.
Score by category
Score and number of tasks per category, on question-answering.
Credit Analysis83.5 · 62
Monitoring97.3 · 20
Structured Products98.6 · 18
Research Intelligence88.3 · 10
Discovery & Prediction98.9 · 2
Where it sits
kimi-k3 against every other model on the board.
kimi-k3 on the best trade-off line — no model beats it on bothBest trade-off · no model beats these on bothBest-by-Task · hypothetical, not a model
On the research track
How it worked when it had to find its own evidence.
Delivered an answer
100%
of research tasks
Average steps
15.4
searches, reads and calculations per task
Where it lost the most
Its lowest-scoring tasks on question-answering.
PARTIAL
Trade Idea Extraction
Research Intelligence
PARTIAL
Spread Analysis
Credit Analysis
PARTIAL
Spread Analysis
Credit Analysis
PARTIAL
Curve Trade Analysis
Credit Analysis
PARTIAL
Initial Credit Assessment
Credit Analysis