IBM

Granite 4.2 3B Instructオープン

IBM Granite 4.2 3B compact SLM optimized for on-device and low-latency inference.

#370リリース日: 2026年7月20日·Open Weights·3 billion
text128K ctx
API providersランキング
Intelligence
42.0 #370
AAQI 36.0
Output speed
148 t/s
TTFT 160 ms · ITL 6.8 ms/tok
Blended price (3:1)
$0.11 /1M
In $0.08/M · Out $0.2/M
Context
128K
≈ 320 A4 pages · max out 8192

Quality Domain Radar

6-Axis Frontier Evaluation vs Category Average & Flagship

Granite 4.2 3B Instruct
Claude Mythos Preview (#1)
Average (68)
Coding (35)Mathematics (38)Hard Science (GPQA) (38)Agentic & Tool Use (40)Knowledge (MMLU) (39)Vision / Multimodal (25)

Radar axes are approximated from category scores where direct benchmark measurements are missing — see 評価方法.

API deployment

Compare hosts & latency

Throughput, TTFT and cache pricing across Groq, Cerebras, Together, DeepInfra, Fireworks, Bedrock and Azure.

GroqCerebrasTogether AIDeepInfraFireworks AIAWS BedrockAzure AI
View providers matrix →

ソース別スコア

モデル詳細

リリース日
2026年7月20日
パラメータ数
3 billion
ライセンス
Open Weights
コンテキストウィンドウ
128,000 tokens
入力価格
$0.08/M
出力価格
$0.2/M
出力速度
148 t/s
最初のトークンまでの時間
160 ms

Estimated monthly bill

Estimated monthly spend
$0.00
Calculating savings vs GPT-4o...

Cite & embed

GitHub Markdown badge
[![LLMPodium Rank](https://llmpodium.com/badge/granite-4-2-3b/rank.svg)](https://llmpodium.com/models/granite-4-2-3b)
BibTeX
@misc{llmpodium2026granite423b,
  title = {Benchmark Analysis and Intelligence Evaluation of Granite 4.2 3B Instruct},
  author = {LLMPodium Research},
  year = {2026},
  url = {https://llmpodium.com/models/granite-4-2-3b}
}

More from IBM

代替モデル

機能、価格、ベンチマークスコアに基づく Granite 4.2 3B Instruct の最適な代替モデル。

0 / 4