Kimi K2.5
moonshotai/kimi-k2.5
Earlier Kimi frontier model for long-context agents, coding, and multimodal work
Capabilities
ReasoningYes
Tool callingYes
Structured outputYes
TemperatureFixed
AttachmentYes
Open weightsYes
Limits
Context262,144 tokens
Max inputN/A
Max output262,144 tokens
Details
Familykimi-k2
Knowledge cutoff2025-01
Release date2026-01
Last updated2026-01
Modalities
Inputtext, image, video
Outputtext
Benchmarks
SWE-Bench Verified
70.8
resolved
SWE-Atlas Codebase QnA
13.1
score
SWE-Atlas Refactoring
20.95
score
SWE-Atlas Test Writing
25.77
score
Community Rating
—
No ratings yet
Sign in to rate
Pricing
Cost per 1M tokens
InputN/A
OutputN/A