Granite 4.2 (3B, Q4)
IBM's compact dense Granite 4.2 reasoning model at 3B parameters. Text-only, with native thinking, native tool calling, multilingual support, and a 128K context window. Q4 builds are a lightweight option for machines with 8GB+ usable memory.
At a glance
- Parameters
- 3B
- Quantization
- Q4_K_M
- Context window
- 131K tokens
- Approx. size
- 2.2 GB
- Engines
- llama.cpp, MLX
- License
- Apache-2.0 (open)
- Version
- v1.0.0 , released 2026-08-25
- Family
- granite
Tuning
Defaults a gezel client applies out of the box, from real evaluation runs against this exact quantization.
- sampling.temperature
1- sampling.topP
0.95- sampling.maxTokens
8192- reasoning.thinkingBudget
8192- reasoning.enableThinking
true- profiles
- thinking-general, thinking-deep, thinking-coding, thinking-precise, instruct, creative
Model behaviors
Client-side behavior modules the catalog enables for this model (reasoning-tag handling, fabrication detection, prompt shaping).
- reasoning.strip-think-tags
- prompt.private-reasoning-guidance
- prompt.tool-cookbook-full
- mcp.compact-tool-schemas
- mcp.relax-required-fields
- mcp.default-missing-fields
- fabrication.detect-past-tense-no-tools
- fabrication.detect-claim-without-tool
- prompt.retrieval-first
- turn.continuation-budget
- turn.preamble-folding
- turn.ramble-detection
- tools.mlx-grammar
- prompt.prefer-writefile-edits
- prompt.derive-by-execution
Sources
- llama.cpp (GGUF)
- ibm-granite/granite-4.2-3b-GGUF ·
granite-4.2-3b-Q4_K_M.gguf· Q4_K_M - MLX
- lmstudio-community/granite-4.2-3b-MLX-4bit
- Upstream
- https://huggingface.co/ibm-granite/granite-4.2-3b
- Manifest
- View on GitHub
Tags
- ibm
- reasoning
- agentic
- coding
- tools
- long-context
- multilingual
- small
- open-weights
- apple-silicon