Granite 4.1 Models by IBM Research
Last refreshed 2026-04-30. Next refresh: weekly.
Details
Capabilities
About
IBM's Granite 4.1 family of dense transformer language models, released April 2026. Available in 3B, 8B, and 30B sizes. Base models support up to 512K token context; instruct models support 131K. Trained on ~15 trillion tokens with improved post-training including SFT and RL alignment for enhanced tool calling and instruction following. Released under Apache 2.0.
Decision facts
- Best fit
- speech recognitionvision and multimodal workcoding
- Capability starting point
- Granite Speech 4.1 2B with 128k context and multimodal inputs
- Lowest tracked input
- Granite 4.1 8B · $0.05/1M · OpenRouter
- Closest related family
- Granite 4
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs 131k context and 3B parameters.
Use when the workload needs 512k context and 3B parameters.
Use when the workload needs 131k context and 8B parameters.
Use when the workload needs 512k context and 8B parameters.
Use when the workload needs 131k context and 30B parameters.
Use when the workload needs 512k context and 29B parameters.
Use when the workload needs speech recognition, 128k context, and 2B parameters.
Use when the workload needs speech recognition, 2B parameters, and multimodal inputs.
Use when the workload needs speech recognition, 2B parameters, and multimodal inputs.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| Granite 4.1 3B | Use when the workload needs 131k context and 3B parameters. | 2026-04 | 131k context3B parameters | Current |
| Granite 4.1 3B Base | Use when the workload needs 512k context and 3B parameters. | 2026-04 | 512k context3B parameters | Current |
| Granite 4.1 8B | Use when the workload needs 131k context and 8B parameters. | 2026-04 | 131k context8B parameters | Current |
| Granite 4.1 8B Base | Use when the workload needs 512k context and 8B parameters. | 2026-04 | 512k context8B parameters | Current |
| Granite 4.1 30B | Use when the workload needs 131k context and 30B parameters. | 2026-04 | 131k context30B parameters | Current |
| Granite 4.1 30B Base | Use when the workload needs 512k context and 29B parameters. | 2026-04 | 512k context29B parameters | Current |
| Granite Speech 4.1 2B | Use when the workload needs speech recognition, 128k context, and 2B parameters. | 2026-04 | speech recognition128k context2B parameters | Current |
| Granite Speech 4.1 2B Plus | Use when the workload needs speech recognition, 2B parameters, and multimodal inputs. | 2026-04 | speech recognition2B parametersmultimodal inputs | Current |
| Granite Speech 4.1 2B NAR | Use when the workload needs speech recognition, 2B parameters, and multimodal inputs. | 2026-04 | speech recognition2B parametersmultimodal inputs | Current |
Release Timeline
1 release groupSpecifications(9 models)
| Model | Released | Context | Parameters | Multimodal |
|---|---|---|---|---|
| Granite 4.1 3B | 2026-04 | 131k | 3B | No |
| Granite 4.1 3B Base | 2026-04 | 512k | 3B | No |
| Granite 4.1 8B | 2026-04 | 131k | 8B | No |
| Granite 4.1 8B Base | 2026-04 | 512k | 8B | No |
| Granite 4.1 30B | 2026-04 | 131k | 30B | No |
| Granite 4.1 30B Base | 2026-04 | 512k | 29B | No |
| Granite Speech 4.1 2B | 2026-04 | 128k | 2B | Yes |
| Granite Speech 4.1 2B Plus | 2026-04 | — | 2B | Yes |
| Granite Speech 4.1 2B NAR | 2026-04 | — | 2B | Yes |
Available From(1 provider)
Pricing
| Model | Provider | Input / 1M | Output / 1M | Type |
|---|---|---|---|---|
| Granite 4.1 8B | OpenRouter | $0.05 | $0.1 | Serverless |
Related Model Families
Models(9)
Granite 4.1 3B
Granite 4.1 3B Base
Granite 4.1 8B
Granite 4.1 8B Base
Granite 4.1 30B
Granite 4.1 30B Base
Granite Speech 4.1 2B
Granite Speech 4.1 2B Plus
Granite Speech 4.1 2B NAR