AGENTKERNEL
- Home
- All models
- Granite 4.0 Micro
Granite 4.0 Micro: price, performance and context
Open Granite 4.0 Micro by IBM costs $0.017 per 1M input and $0.112 per 1M output tokens, handles up to 131K tokens in one request (≈ 390 pages of text) and scores 5.1 points on the AgentKernel performance index, rank 207 of 212.
Privacy: no zero-retention provider listed.
Open in the interactive comparisonKey facts
- Licence
- Open weights: free to download and self-host
- Input / 1M tokens
- $0.017
- Output / 1M tokens
- $0.112
- Context window
- 131K tokens
- Max. output
- 117.9K tokens
- Performance index
- 5.1 points · rank 207 of 212
- Coding index
- n/a
- Agentic index
- n/a
- Cost at your usage
- $0.045 / month
- Knowledge cutoff
- n/a
- In the catalog since
- Oct 20, 2025
- Capabilities
- Text
What do you get for $10.00?
Rules of thumb from the cost calculator, base rates without caching or taxes.
| Task | Cost per task | For $10.00 |
|---|---|---|
| Chat repliesA question with a short history, a detailed answer | $0.00 | 164,000 |
| Email draftsBullet points in, a finished email out | $0.00 | 204,000 |
| Document summariesRead 10 pages, write a one-page summary | $0.00 | 89,000 |
| Code reviewsReview and comment on 500 lines of code | $0.00 | 49,000 |
| Book analysesAnalyse a 300-page book in one go | $0.002 | 5,200 |
Cheaper with similar performance
| Model | Licence | Performance | Your cost / month |
|---|---|---|---|
| Ling 3.1 FlashinclusionAI | Proprietary | 41.1 | $0.00 |
| Ling 3.0 Flash VLinclusionAI | Open | 24.6 | $0.036 |
| Ling 3.0 FlashinclusionAI | Open | 20.1 | $0.037 |
Stronger models
| Model | Licence | Performance | Your cost / month |
|---|---|---|---|
| Ministral 3 8B 2512Mistral | Open | 5.5 | $0.188 |
| GPT-3.5 TurboOpenAI | Proprietary | 5.5 | $0.875 |
| Llama 3.2 3B InstructMeta | Open | 5.7 | $0.133 |