Models · IBM Granite · Out sinceReleased 25 Aug 2026

Granite 4.2 8B

Available on OpenRouterOn OpenRouterAvailable nowGenerally availableCan run on your own serversOpen weights

Granite 4.2 8B is made by IBM Granite. We don't have enough test results yet to rank it. It's cheap to use. Its makers have published it, so you can run it on your own servers.

Dense 8B reasoning model, Apache 2.0, built on Granite-4.1-8B-Base. Native 128K context (extension to 512K). OpenRouter ~$0.06/$0.25.

Writing & creativity: —
Research & analysis: —
Coding: —
Price: Cheap$0.06 / $0.25
Price per 1M (blended)Blended / 1M: $0.11
MemoryContext: 131K
Longest answerMax output: —
Test resultsResults: 8 (8 independent8 indep.)
Out sinceReleased: 25 Aug 2026
Made byVendor: IBM Granite
UnderstandsInputs: text

Scores are out of 100 for each area, compared with every model we track; “#” is its rank. Context is how much text it can read at once (1M is roughly 700,000 words). Blended price mixes the cost of what you send and what it writes back.

Thinking levels

Reasoning settings

No reasoningLowDefault

Links

Using it through OpenRouter

On OpenRouter

OpenRouter is a service that gives access to many AI models in one place. This model is listed there as ibm-granite/granite-4.2-8b.

Route it as ibm-granite/granite-4.2-8b at $0.06 in / $0.25 out per 1M tokens, 131K context. Listed since 31 Aug 2026.

frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningreasoning_effortrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p

Run it yourself

Self-hosting

On your own servers,Licence, sizeunder your own control.and hardware.

Compare all open models →All open-weights models →

LicenceLicence: Apache-2.0
OK for business useCommercial use: Yes
SizeParams (total / active): 8B
Can be trained furtherBase model: Yes ↗

Hardware you'd need

Hardware tier & memory

Runs on a powerful laptop or workstation.

≤ 24 GB at 4-bit (one consumer GPU / 32 GB Mac). Weights ≈ 8 GB at 8-bit, 4 GB at 4-bit (+10–30% for KV cache).

Languages: Languages: Same 12 tested languages as 30B (no Nordic).

Training it further: Fine-tuning: No recipe in card; base Granite-4.1-8B-Base published.

Quantisations: bf16, fp8, nvfp4, mxfp4, gguf, mlx. Engines: vLLM, SGLang, Transformers.

Download (Hugging Face)Weights on Hugging Face ↗huggingface.co ↗

Every test result

Every result

All the numbers,with where they came from.

Every result we've found for this model, with who measured it and a link to where we read it.

All raw rows (vendor and independent kept separate), with setting label, source, date, harness and notes.

All results for this model
						NotesNotes
AA-LCR	45.0%	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b
AA-Omniscience Index	-17.2	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b; AA-Omniscience Index (-100..100)
Artificial Analysis Intelligence Index	11.1	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b
Artificial Analysis output speed	66 tok/s	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b; median output tokens/sec across AA-tracked API providers (not a self-hosted measurement)
GPQA Diamond	63.1%	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b
Humanity's Last Exam	9.7%	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b
SciCode	31.5%	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b
Terminal-Bench 2.1	18.4%	DefaultGranite 4.2 8B	Independent testIndependentArtificial Analysis ↗	1 Oct 2026	—	AA slug granite-4-2-8b