Granite 4.2 30B
Granite 4.2 30B is IBM's flagship dense reasoning model in the Granite 4.2 family released August 25, 2026.
Provider
IBM
Model family
IBM Granite
Open-weight reasoning LLM
Cost tier
4 2 30b
Status
Current
Release Aug 25, 2026
Why teams choose it
IBM released Granite 4.2 in 3B, 8B, and 30B dense sizes
this catalog row is the 30B flagship.
Every size includes a thinking / non-thinking switch
IBM says the 8B and 30B models also received agentic RL in sandboxed tool environments.
IBM documents a 128K native context
with 512K long-context extension on the 30B model.
Tradeoffs to know
- Vendor coding and agent claims should be reproduced against your tools, policies, and evaluation harness.
- IBM also announced Granite Speech 5.0 Turbo CTC models in the same post; those are separate speech checkpoints, not this LLM.
When not to use this
- The 3B and 8B siblings are better fits for edge or high-throughput serving; do not assume 30B quality at those sizes.
Technical specs
- Inputs
- text
- Outputs
- text
- Capabilities
- reasoning, tool calling, coding agents, enterprise agents, thinking mode, on-prem deployment
- License
- Apache 2.0
- Source
- https://research.ibm.com/blog/introducing-granite-4-2
- Parameters
- 30B
- Native context
- 131,072
Benchmarks
No independent or vendor benchmark scores in this catalog entry.
Granite 4.2 30B FAQ
What is Granite 4.2 30B?
Granite 4.2 30B is IBM's flagship dense reasoning model in the Granite 4.2 family released August 25, 2026. IBM documents Apache 2.0 weights in 3B, 8B, and 30B sizes, a native thinking/non-thinking switch, native tool calling, a 128K context window, and long-context extension to 512K on the 30B model.
When does Granite 4.2 30B fit best?
On-prem enterprise agents
What should teams watch out for with Granite 4.2 30B?
The 3B and 8B siblings are better fits for edge or high-throughput serving; do not assume 30B quality at those sizes.
Explore next
Models, tools, and comparisons that connect to this reference.