by Moonshot AI
Kimi K3 reviews
Moonshot AI's flagship model. It's built for agent work and long, multi-step tasks, with strong coding scores.
$15 per 1M output tokens · Intelligence 43.6
No sign-up needed. Just confirm your email.
Reviews
Be the first to review Kimi K3
Used it? Tell people what it did well and where it let you down. Your review helps everyone choose.
No sign-up needed. Just confirm your email.
How Kimi K3 compares on price
Kimi K3 scores 43.6 at $15 per 1M output tokens. GPT-6 Astra leads with 52.7. GLM 5.3 Flash gets 96% of its score for 1/30 of the price.
What changed since Kimi K2.6
Kimi K3 is the newer release. Reviews of Kimi K2.6 stay on its own page.
| Kimi K2.6 | Kimi K3 | Change | |
|---|---|---|---|
| Price / 1M output | $4 | $15 | +275% |
| Intelligence | 27 | 43.6 | +16.6 |
| Agentic | 20.5 | 50 | +29.5 |
| Coding | 61.8 | 76.2 | +14.4 |
| $5 free credit | ≈ 740 replies | ≈ 200 replies | −73% |
Make up your own mind about Kimi K3
Your free $5 covers about 200 typical replies from Kimi K3, or roughly 250,000 words of writing. Run everyday work on GLM 5.3 Flash and it stretches to about 5,900. No card needed.
How we estimate: a typical reply is about 550 words, written with about 3,750 words of conversation sent along with it, at Saaga's billed price for Kimi K3. Long documents, file uploads and agents that call tools use more per reply; quick questions use less. “Words of writing” counts output only.
Compare with other models
Try Kimi K3 freeYour $5 covers ≈ 200 replies
Try it