OMMAIS Blog · Writers

OMMAIS: Mistral Nemo

Made by Mistral AI and NVIDIA · released July 2024

A small, fast 12-billion-parameter model with a long 128,000-token memory. It is the free default in OMMAIS Chat because it answers quickly and reliably.

Long OMMAIS Chat's free default model, used when no other provider is set up.

Good at

• Quick, clear explanations
• Summarising long material
• Many languages

Published benchmarks

• 68.0% MMLU (base model) general knowledge across 57 subjects
• 83.5% HellaSwag (base model) common-sense reasoning about everyday situations
• 76.8% WinoGrande (base model) resolving ambiguous sentences by common sense

Scores as reported by the model’s makers. Benchmarks measure narrow skills.

Use it yourself in OMMAIS Chat

1. Get a key from one of the providers below (the link opens their key page). Some need no key at all.
2. Open OMMAIS Chat, go to Settings → Cloud providers, and paste the key next to that provider.
3. Pick Mistral Nemo (or the closest current version) from that provider's model list. The lists are read live from each provider.

• L7 LLM7.io · free and paid models No key needed; a free key raises the limit. No key needed.
• M Mistral · free tier Mistral's own free experiment plan. Get a Mistral key →

The Mistral AI and NVIDIA mark is a trademark of its owner, shown only to identify the model.