Models / Compare /DeepSeek V3.2 vs Kimi K3
Head to head
DeepSeek V3.2 vs Kimi K3
Compared on live catalog data — pricing synced daily, benchmark scores with provenance.
The short version
- DeepSeek V3.2 is cheaper to run hosted: $0.40/M output vs $15.00/M for Kimi K3 (37.5× difference at the cheapest tracked provider).
- Kimi K3 takes 1M tokens of context vs 164K for DeepSeek V3.2.
- DeepSeek V3.2 has downloadable weights (OPEN) so you can self-host it; we hold no downloadable copy of Kimi K3, so it runs through a provider's API.
- Kimi K3 scores 1531.4745 via Max on Arena Coding vs 1469.5929 for DeepSeek V3.2.
- Kimi K3 scores 1468.7657 via Max on Arena Creative Writing vs 1400.3738 for DeepSeek V3.2.
| Spec | ||
|---|---|---|
| Vendor | DeepSeek | Moonshot AI |
| Parameters | 685B / 37B | 2.8T / 104B |
| 164K | 1M | |
| Weights | OPEN | OPEN* |
| License | MIT License | Custom licence |
| Modality | text->text | text+image->text |
| Released | Dec 1, 2025 | Jun 13, 2026 |
| 432.1 GBest | 1752.7 GBest | |
| Cheapest hosted | $0.27 / $0.40 per 1M (Novita AI) | $3.00 / $15.00 per 1M (Novita AI) |
| Tracked providers | 19 | 17 |
DeepSeek V3.2 — our take
DeepSeek V3.2 is a large downloadable text model with a permissive MIT licence and a 163,840-token request limit. Only 37 billion of its 685.4 billion total parameters are active per token, keeping inference costs down while measured coding performance is the clear peak in its benchmark profile.
full breakdown →Kimi K3 — our take
Kimi K3 is Moonshot AI's flagship coding model with nearly 2.8 trillion parameters and a one-million-token request limit. Its measured coding performance sits at the frontier, though its custom licence carries restrictions and its agentic scores are thinly measured.
full breakdown →Need more columns? Use theinteractive comparison to add up to four models.