AI Models
Compare the available AI models and learn when to use each one.
Available models
Localizia offers ten AI models for translation from three providers, ranging from fast and affordable to premium quality.
Budget tier
| Model | Provider | Best for | Input (per 1K tokens) | Output (per 1K tokens) |
|---|---|---|---|---|
| Gemini 2.5 Flash-Lite | Cheapest option, high-volume simple content | 1 credit | 4 credits | |
| GPT-5.4 Nano | OpenAI | Ultra-fast, short strings and labels | 2 credits | 9 credits |
| Claude Haiku 3 | Anthropic | Fast and affordable general content | 3 credits | 13 credits |
| Gemini 2.5 Flash | Fast with good quality balance | 3 credits | 18 credits | |
| GPT-5.4 Mini | OpenAI | Versatile budget option | 6 credits | 32 credits |
| Claude Haiku 4.5 | Anthropic | Smart budget option, best quality in tier | 7 credits | 35 credits |
Standard tier
| Model | Provider | Best for | Input (per 1K tokens) | Output (per 1K tokens) |
|---|---|---|---|---|
| Gemini 2.5 Pro | High quality at moderate cost | 9 credits | 70 credits | |
| GPT-5.4 | OpenAI | General high-quality content | 18 credits | 105 credits |
| Claude Sonnet 4.6 | Anthropic | Marketing copy, documentation | 21 credits | 105 credits |
Premium tier
| Model | Provider | Best for | Input (per 1K tokens) | Output (per 1K tokens) |
|---|---|---|---|---|
| Claude Opus 4.6 | Anthropic | Technical, legal, nuanced content | 25 credits | 125 credits |
Choosing the right model
Budget models
Best for high-volume translations where cost matters most:
- App UI strings and labels
- Simple notification messages
- Content that will be reviewed and edited by a human
- Large batches of short texts
Standard models
A balanced choice for most use cases:
- Blog posts and articles
- Product descriptions
- Marketing copy and landing pages
- Customer-facing documentation
Premium model (Claude Opus 4.6)
The best quality available, ideal for:
- Technical documentation and manuals
- Legal and regulatory content
- Medical and scientific texts
- Any content requiring precise terminology and nuanced understanding
How costs are calculated
Translation costs depend on two factors:
- Text length — measured in tokens (roughly 4 characters per token)
- Model selected — premium models cost more per token
The cost formula:
credits = (input tokens × input rate) + (output tokens × output rate)You always see the estimated cost before confirming a translation. The actual cost may vary slightly depending on how the AI structures its response.
Default model
New projects default to GPT-5.4 Nano. You can change the model each time you start a translation job — different jobs in the same project can use different models.
Tips
- Start with a budget model to test the workflow, then upgrade if you need higher quality
- Use Claude Opus 4.6 for critical content that won't be heavily edited
- For large files with mixed content, consider translating in batches with different models — use a budget model for simple strings and a premium model for complex paragraphs