$ diff grok-4-6 kimi-k3
Grok 4.6 vs Kimi K3
Values come from Grok 4.6 and Kimi K3, where each is cited to its source. This page states no benchmark result and ranks neither model — it puts two published specifications next to each other. Where a lab has not published a figure, the row says so rather than guessing.
xAIGrok 4.6
- context
- 500K
- weights
- closed
- $/M in
- $2
- $/M out
- $6
- context
- 1M
- weights
- open
- $/M in
- $3
- $/M out
- $15
What actually differs
- Context window
- Kimi K3 takes 1M against 500K — 2× more room in a single request.
- Input price
- Grok 4.6 at $2/M against $3/M — 1.5× cheaper to feed. Standard rates: one of these labs also quotes a lower cached-input tier, which applies only when a prefix is reused — see the full spec below.
- Output price
- Grok 4.6 at $6/M against $15/M — 2.5× cheaper to generate. Output dominates the bill on most agentic workloads, where the model writes far more than it reads.
- Weights
- Kimi K3 publishes weights (Modified MIT (commercial use permitted; weights publicly downloadable)); Grok 4.6 is API-only. That decides self-hosting, air-gapped deployment and fine-tuning before any capability question does.
- Recency
- Grok 4.6 shipped 22 days after Kimi K3 (2026-08-07 vs 2026-07-16).
Full spec
| Attribute | Grok 4.6 | Kimi K3 |
|---|---|---|
| Developer | xAI | Moonshot AI |
| Released | 2026-08-07 | 2026-07-16 |
| Context window | 500,000 tokens | 1M tokens |
| Pricing | $2/M input · $6/M output; a request reaching 200K tokens re-prices in full at $4.00/M input · $1.00/M cached input · $12.00/M output | $0.30/M cache-hit input · $3.00/M cache-miss input · $15.00/M output |
| License | proprietary | Modified MIT (commercial use permitted; weights publicly downloadable) |
| Availability | API | Kimi Code, Kimi app, Kimi API; weights on Hugging Face (`moonshot-ai/kimi-k3`, released 2026-07-27) |