$ diff kimi-k3 muse-spark
Kimi K3 vs Muse Spark
Moonshot AI and Meta AI, side by side. Every value below is the one recorded on the model’s own wiki page, with its citation — nothing is estimated.
Moonshot AIKimi K3
- context
- 1M
- weights
- open
- $/M in
- $3
- $/M out
- $15
- context
- 1M
- weights
- open
- $/M in
- $1.25
- $/M out
- $4.25
What actually differs
- Context window
- Identical — both accept 1M tokens, so context length is not a reason to pick either.
- Input price
- Muse Spark at $1.25/M against $3/M — 2.4× 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
- Muse Spark at $4.25/M against $15/M — 3.5× cheaper to generate. Output dominates the bill on most agentic workloads, where the model writes far more than it reads.
- Weights
- Both publish weights — Kimi K3 under Modified MIT (commercial use permitted; weights publicly downloadable), Muse Spark under proprietary (closed / API-only, not open-weights). Check the licences rather than assuming they permit the same commercial use.
- Recency
- Kimi K3 shipped 99 days after Muse Spark (2026-07-16 vs 2026-04-08).
Full spec
| Attribute | Kimi K3 | Muse Spark |
|---|---|---|
| Developer | Moonshot AI | Meta AI |
| Released | 2026-07-16 | 2026-04-08 |
| Context window | 1M tokens | 1,000,000 tokens (1.1) |
| Pricing | $0.30/M cache-hit input · $3.00/M cache-miss input · $15.00/M output | $1.25/M input · $4.25/M output (Meta Model API, public preview) |
| License | Modified MIT (commercial use permitted; weights publicly downloadable) | proprietary (closed / API-only, not open-weights) |
| Availability | Kimi Code, Kimi app, Kimi API; weights on Hugging Face (`moonshot-ai/kimi-k3`, released 2026-07-27) | Meta Model API (public preview, US-based developers only) |
Values come from Kimi K3 and Muse Spark, 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. How these pages are produced.