Model

Granite 3.1 8B Instruct

OPEN

IBM · Granite 3.1 family · released Dec 18, 2024

Extends Granite 3.0's context from 4K to 128K via progressive RoPE training on ~500B additional tokens.

ReasoningCodingVisionFunction callingTool useAgentic

371.2

Elo · rank #339

Parameters

8.1B

Active params

8.1B (dense)

Context

128K tokens

Architecture

Dense decoder-only transformer (GQA, RoPE with adjusted theta for 128K context)

License

Apache 2.0

Languages

12+

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

ARC-ChallengeReasoning62.6%#76

best: Llama 3.1 405B · 96.9%

BIG-Bench HardReasoning34.1%#120

best: ERNIE 4.5 300B-A47B · 94.3%

GPQA DiamondReasoning8.3%#286

best: GPT-5.6 · 94.6%

GSM8KMath73.8%#95

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning84.5%#47

best: Claude 3 Opus · 95.4%

IFEvalReasoning72.1%#110

best: Gemma 4 26B A4B · 98.5%

MATH-500Math21.7%#180

best: GPT-5 · 99.4%

MMLU-ProKnowledge28.2%#165

best: Claude Fable 5 · 91.5%

MMLUKnowledge65.3%#176

best: OpenAI o3 · 92.9%

TruthfulQAKnowledge66.2%#13

best: Phi-3.5-MoE (16x3.8B, 6.6B active) · 77.5%

WinoGrandeReasoning75.4%#62

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

6 GB

VRAM @ FP16

16.2 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4_K_M · AWQ · GPTQ

Granite 3.1 family

Elo progression across releases

Granite 3.1 8B InstructDec 2024371.2

Resources

HFibm-granite/granite-3.1-8b-instruct GHibm-granite/granite-3.1-language-models DOCSibm.com/granite/docs/models/granite APIAvailable via IBM API

API price weights · each benchmark row carries its own source badge (see methodology)