Currently listed through these providers:
Model details
Granite 4.2 8B
Granite 4.2 8B sits in the middle of IBM's Granite 4.2 family, a set of dense reasoning language models released in 3B, 8B, and 30B sizes. The 8B variant is positioned as a balanced option for general-purpose enterprise workloads, sitting between a compact 3B model aimed at edge and resource-constrained deployments and a 30B flagship tuned for the most complex reasoning. Across the family, IBM introduced native chain-of-thought reasoning so the models work through problems step by step before answering, with flexible thinking modes that can be tuned for different tasks and reasoning-augmented tool calling that helps the model plan and verify calls to external functions.
The architecture is a dense transformer-style language model rather than a mixture-of-experts design, and the 8B size makes it practical to self-host or fine-tune while still benefiting from the family's reasoning improvements. According to IBM, Granite 4.2 models are built to handle complex math, coding, multi-step logic, and agentic tool-calling scenarios, where breaking a problem into intermediate steps and using external tools are central. The 8B variant is a good fit for teams that want open weights they can run internally for cost or data-control reasons, while still getting the step-by-step reasoning behavior that distinguishes the Granite 4.2 generation from earlier IBM releases.
Quick Info
Powered by- Provider
- OpenRouter
- Model key
- ibm-granite/granite-4.2-8b
- Release date
- Aug 31, 2026
- Last updated
- Aug 31, 2026
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.10
- Output token cost
- $0.15
Limits
- Output tokens
- 117,964 tokens
- Context window
- 131,072 tokens