Granite 4.2 8B
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Instruction following#4286.5
- 2Knowledge58.6
- 3Reasoning55.9
- 4Coding#12518.5
Benchmark results (13)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Coding
| SWE-bench Verified | 47.7 | #88 of 98 |
| SWE-bench Pro | 19.1 | #68 of 69 |
| LiveCodeBench v6 | 73.2 | #30 of 46 |
| SciCode | 36.1 | #18 of 23 |
Agentic & tool use
| Terminal-Bench 2.1 | 20.6 | #66 of 68 |
| τ³-bench results | 58.1 | #14 of 23 |
| BFCL v4 | 52.4 | #12 of 19 |
Knowledge
| GPQA | 64.1 | #88 of 117 |
| MMLU-Pro | 74.0 | #48 of 67 |
Math
| AIME 2025 | 86.7 | #18 of 38 |
| HMMT Feb 2025 | 78.3 | #12 of 13 |
Multilingual
| SWE Multilingual | 30.8 | #45 of 46 |
Instruction following
| IFBench | 79.3 | #13 of 42 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| List price (OpenRouter) | $0.060 | $0.25 | — |
| OpenRouter ibm-granite/granite-4.2-8b | $0.060 | $0.25 | $0.015 |
IBM publishes Granite 4.2 8B under Apache 2.0 for self-hosting and does not publish a first-party hosted token rate for this exact checkpoint. We represent the open-weight row as self-host/free-per-token before infrastructure costs. OpenRouter currently exposes ibm-granite/granite-4.2-8b as a hosted route, but its third-party rate is not used as canonical pricing.
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| Text | 1320 | #234 · 3.3K votes |
Open weights
From the Hugging Face repository.
- Parameters
- 8.8B
- Licence
- apache-2.0
- Downloads (30 days)
- 90.5K
- Repository
- ibm-granite/granite-4.2-8b ↗
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Same family
- Granite-4.0-H-1B28.8
Similar score
- GPT-OSS 20B33.3 · $0.036
- Kimi K233.5 · $1.07
- Mistral Large 333.5 · $0.75
- Mistral Medium 3.5 128B33.6 · $3.00
- Mercury 233.7 · $0.38