Kimi K3
The highest-placed open-licence model on Arena, and the one most teams have not tried yet.
Key facts
- Score
- 60 / 100, rank #11
- Price
- $2.31 per 1M blended (Artificial Analysis estimate)
- Blended price
- $2.31 per 1M tokens (3 input : 1 output), estimated
- Context window
- Not verified
- Open weights
- Yes
- Best for
- Self-hosting, Open-licence requirements, Arena-style chat quality
Moonshot does not publish a simple list price for Kimi K3; Artificial Analysis estimates a blended $2.31 per million tokens (three input to one output).
Independent evals
| Benchmark | Result | Who ran it | Read |
|---|---|---|---|
| Arena text score (max) | 1488 | Arena text leaderboard (Oct 8, 2026) | Oct 9, 2026 |
| Artificial Analysis Intelligence Index (max) | 44 | Artificial Analysis LLM leaderboard | Oct 10, 2026 |
- Artificial Analysis blended price (max): $2.31 per 1M (Artificial Analysis LLM leaderboard)
Caveats
- Price is Artificial Analysis' blended figure, not a vendor list price; context not verified from a primary source.
Questions people ask
How much does Kimi K3 cost?
Moonshot does not publish a simple list price for Kimi K3; Artificial Analysis estimates a blended $2.31 per million tokens (three input to one output).
How good is Kimi K3?
Kimi K3 ranks #11 of 12 on the Models at Work leaderboard with a score of 60 out of 100 (October 2026 edition). The highest-placed open-licence model on Arena, and the one most teams have not tried yet.
What is Kimi K3 best for?
Kimi K3 is best for self-hosting, open-licence requirements, arena-style chat quality.
What are Kimi K3's benchmark scores?
Arena text score (max): 1488 (Arena text leaderboard (Oct 8, 2026)); Artificial Analysis Intelligence Index (max): 44 (Artificial Analysis LLM leaderboard).
Is Kimi K3 open weights?
Yes. Kimi K3 is released with open weights, so it can be self-hosted.
Compare Kimi K3
Methodology: Score is 0 to 100: 50% independent evals normalised within this table, 20% price-performance, 15% practitioner sentiment from Signals and community threads, 15% operational fit (context, speed, availability, open weights). Models within three points are a tie in practice. Vendor-published figures are labelled as such and never counted as independent.