IBM Granite 3.2
Granite 3.2
Boring, Apache-licensed, cryptographically signed, and built for a procurement review.
Updated
The read
Granite is aimed at the enterprise buyer rather than the benchmark table: Apache 2.0, small enough to run cheaply, with signed releases and a documented data provenance story that a compliance team can actually read. For classification, summarising, extraction and code assistance inside a regulated organisation, that paperwork is worth more than a few benchmark points. Outside that context there are stronger models at the same size.
Where it shines
- Cryptographically signed releases with documented provenance
- Apache 2.0 with indemnification available through IBM
- Small enough to run on ordinary enterprise hardware
Worth knowing first
- Outclassed on capability by open rivals of the same size
- Short context windows across the family
Index score
72.5
Provisional. This score is read from the product's public capability, not from the one-prompt rebuild the fully reviewed entries went through, so treat it as a placing rather than a verdict.
- Rank
- 44 of 53
- Builds
- LLMs
- Category
- Open weights
- Output
- Apache-licensed weights, signed releases
- Pricing
- Free tier
- Best for
- Regulated organisations that must document what they deploy
Scores are our own editorial judgement, weighted craft 55, speed 10, control 15 and value 20.
Also worth a look
Full index- 05Kimi K3Kimi K3ProvisionalThe largest open-weights model anyone has shipped, and it reads a million tokens without blinking.Open weightsFree tier87.5
- 06GLM-5.2GLM-5.2ProvisionalFrontier-class output, an MIT licence, and a million tokens of context. That combination is the story.Open weightsFree tier87.1
- 12Qwen 3.6Qwen 3.6ProvisionalThe open Qwen release most people should actually run, in sizes that fit real hardware.Open weightsFree tier83.3
