GPT-5.4 Mini vs Nemotron 3 Ultra: cost per task
GPT-5.4 Mini is cheaper on 12 of 14 tasks and about 2.1x cheaper on average (geometric mean). GPT-5.4 Mini has the higher LMArena Text score (1447 vs 1426).
| Task | GPT-5.4 Mini | Nemotron 3 Ultra | Difference |
|---|---|---|---|
| Answer a chat question | 0.53¢ | 0.33¢ | 1.6x |
| Reply to a customer email | 0.06¢ | 0.86¢ | 14.0x |
| Write a 1,500-word blog post | 1.1¢ | 1.9¢ | 1.7x |
| Write 10 product descriptions | 0.48¢ | 0.75¢ | 1.6x |
| Translate 1,000 words | 0.73¢ | 3.0¢ | 4.1x |
| Proofread one page (500 words) | 0.39¢ | 1.0¢ | 2.6x |
| Summarize a 20-page PDF | 1.2¢ | 1.3¢ | 1.1x |
| Extract data from an invoice | 0.28¢ | 0.66¢ | 2.3x |
| Analyze a 200-row spreadsheet | 1.6¢ | 2.8¢ | 1.8x |
| Answer from 5 retrieved documents (RAG) | 0.43¢ | 0.99¢ | 2.3x |
| Classify 100 support tickets | 0.56¢ | 1.8¢ | 3.3x |
| Review 500 lines of code | 2.7¢ | 2.5¢ | 1.1x |
| Write a SQL query | 0.36¢ | 0.76¢ | 2.1x |
| Run a 50-step coding agent | 32¢ | 67¢ | 2.1x |
Specs and list prices
| GPT-5.4 Mini | Nemotron 3 Ultra | |
|---|---|---|
| Input per 1M | $0.75 | $0.60 |
| Output per 1M | $4.50 | $2.40 |
| Cached input per 1M | $0.075 | $0.12 |
| Context | 400K | 262K |
| Providers | 5 | 5 |
| Reasoning | default | default |
| LMArena Text | 1447 | 1426 |
| Released | Mar 17, 2026 | Jun 4, 2026 |
FAQ
Is GPT-5.4 Mini or Nemotron 3 Ultra cheaper?
GPT-5.4 Mini is cheaper on 12 of 14 tasks and about 2.1x cheaper on average (geometric mean). GPT-5.4 Mini has the higher LMArena Text score (1447 vs 1426).
How much does GPT-5.4 Mini cost compared to Nemotron 3 Ultra per token?
GPT-5.4 Mini: $0.75 input and $4.50 output per 1M tokens. Nemotron 3 Ultra: $0.60 input and $2.40 output. Token prices alone mislead - models use different numbers of tokens for the same task, which is what the task table shows.