Modelverse keeps model facts separate from Run evidence and warns whenever benchmark conditions are not equivalent.
These models do not currently have leading Runs under identical evaluation conditions. Values are shown with their provenance and should not be treated as a universal leaderboard.
| Attribute | Nova Instruct 12B community/nova-instruct-12b | Sparrow Chat 3B local-ai/sparrow-chat-3b |
|---|---|---|
| Openness | Open weights | Open weights |
| Commercial use | Permitted | Permitted |
| Parameters | 12B | 3.2B |
| Context length | 33K tokens | 16K tokens |
| Estimated memory | 8.2 GB · Q4_K_M | 2.8 GB · Q4_K_M |
| Runtimes | Ollama, llama.cpp, Transformers, vLLM | Ollama, llama.cpp, MLX |
| Tasks | Chat, Summarization, Reasoning, Translation | Chat, Summarization, Extraction |
| Verified Runs | 9 | 16 |
| Best evidence | Machine-captured | Independently reproduced |
| Task score | 71.3 Multilingual instruction suite | — |
| Speed | 118.4 tok/s | 18.9 tok/s |
| Peak memory | 18.8 GB | 3.5 GB |
| Run hardware | RTX 3090 · 24 GB | Apple M1 · 8 GB |
A general-purpose multilingual instruction model with broad runtime support.
A fast, memory-efficient chat model for laptops and edge devices.
First: eliminate models that fail the license, task, modality or runtime requirement.
Second: confirm fit on equivalent hardware or preserve a generous estimated memory margin.
Third: compare only Runs sharing an evaluation suite, revision and broadly equivalent configuration.
