Kimi-K3-DSpark vs Llama-3.2-3B-Instruct
Specs, VRAM requirements and download trends — updated 24 October 2026.
meta-llama · B · Open weights
Params
3.2 B
Context
—
30d
1.1 M
llama3.2 licence · conditional
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | Kimi-K3-DSpark | Llama-3.2-3B-Instruct |
|---|---|---|
| Deployment | Open weights | Open weights · also via API |
| Parameters | 2.3 B | 3.2 B |
| Architecture | DSparkDraftModel | LlamaForCausalLM |
| Context length | 1,048,576 | 80,000 |
| Licence | — | llama3.2 |
| Commercial use | Unknown | Conditional |
| Languages | — | 8 |
| Downloads 30d | 3,725,706 | 1,112,418 |
| Downloads all time | 11.0 M | 50.4 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | No | Yes — accept licence |
| First seen on the Hub | 2026-07-27 | 2024-09-18 |
| API input price / 1M tokens | not on OpenRouter | $0.051 |
| API output price / 1M tokens | not on OpenRouter | $0.34 |
| Est. cost per 1M input + 1M output tokens | not on OpenRouter | $0.39 |
| Cached input / 1M tokens | not on OpenRouter | — |
| AA Intelligence Index | not on OpenRouter | 5.7 |
| AA Coding Index | not on OpenRouter | — |
| AA Agentic Index | not on OpenRouter | — |
| Max output tokens | not on OpenRouter | 72,000 |
| API providers | not on OpenRouter | 2 |
| Tool calling | not on OpenRouter | No |
| Reasoning mode | not on OpenRouter | No |
API cost
Estimated cost of 1M input + 1M output tokens via OpenRouter
Llama-3.2-3B-Instruct
$0.39
Download trend
Daily snapshots, last 84 days (2 Aug – 24 Oct)
6.6 M168 K
Kimi-K3-DSpark
Llama-3.2-3B-Instruct
A overtook B on 10 Aug
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
Llama-3.2-3B-Instruct · fp16
8.0 GB / 24 GB
Kimi-K3-DSpark · fp16 @ 1024K ctx
48.6 GB / 24 GB