Models
The Assistant model picker lets you choose which foundation model drives each conversation. Nomic currently supports four models. All prices below are passed through at the provider's API rate and are quoted per 1M tokens.
Picking a model
- Sonnet 4.6 — Recommended default. Best balance of speed, cost, and quality for daily Assistant use and most workflows.
- Opus 4.6 — Use for the hardest reasoning, ambiguous specs, and high-stakes reviews. Roughly 1.7× Sonnet's cost.
- Haiku 4.5 — Use for quick lookups, exploration, and high-volume automation. Roughly 3× cheaper output than Sonnet.
- Gemini 3.1 Pro — Use when you need very large context windows, or when your team prefers a Google-based inference path.
Switching models inside a session does not change how Assistant interacts with your files — the same tools, citations, and context apply to every model.
Pricing
| Model | Provider | Input | Output | Cache write | Cache read |
|---|---|---|---|---|---|
| Haiku 4.5 | Anthropic | $1.00 | $5.00 | $1.25 | $0.10 |
| Sonnet 4.6 | Anthropic | $3.00 | $15.00 | $3.75 | $0.30 |
| Opus 4.6 | Anthropic | $5.00 | $25.00 | $6.25 | $0.50 |
| Gemini 3.1 Pro | $2.00 | $12.00 | — | $0.20 |
Prices are USD per 1M tokens. Anthropic models charge separately for cache writes; Gemini does not. Cache reads on long multi-turn threads are typically 10× cheaper than fresh input — Assistant takes advantage of this automatically when you continue a conversation in the same session.
Cost in practice
Using the recommended default (Sonnet 4.6):
- A single short Assistant turn typically costs ~$0.02 – $0.05.
- A multi-message thread with ~10 tool calls and a couple thousand output tokens costs ~$0.10 – $0.25.
- A Deep Research run across a large project can reach ~$0.50 – $1.50.
For a side-by-side comparison of how far a given AI Usage allocation goes across real workflows, see the Models & Pricing overview.