Granite 4.2 30B
IBM開源權重Apache 2.0 · 商用許可
描述
IBM Granite 4.2 30B is the flagship dense decoder-only reasoning LLM in the Granite 4.2 family. It supports native think, non-think, and low-effort modes, tool calling, and a 128K native context window with extension to 512K. The 8B and 30B sizes received additional agentic reinforcement learning for multi-step tool use. Open weights on Hugging Face under Apache 2.0. Text-only (not multimodal).
發布日期
2026-08-25
參數規模
30.0B
上下文長度
—
支援模態
text
能力雷達圖
12
general
31
coding
64
reasoning
44
science
50
agents
0
multimodal
排行榜排名
基準測試分數 (LLM Stats)
(LLM Stats (zeroeval))Agents
TAU3-Bench
62.0%自報
General
Arena-Hard v2
67.9%自報
Instruction Following
IFBench
77.2%自報
Language
MMLU-Pro
77.6%自報
Long Context
RULER 64k
90.0%自報
RULER 128k
81.4%自報
Math
HMMT25
89.2%自報
AIME 2025
89.2%自報
Reasoning
LiveCodeBench v6
75.8%自報
GPQANYU + Cohere + Anthropic (2023)
66.4%自報
SWE-Bench Verified
57.0%自報
ProfBench
42.9%自報
SWE-bench Multilingual
41.9%自報
Bird-SQL (dev)
41.9%自報
SciCode
38.8%自報
SWE-Bench ProPrinceton NLP (2024)
33.3%自報
Terminal-Bench 2.1
29.2%自報
Tool Calling
BFCL-V4
61.4%自報
AA 評測指數
(Artificial Analysis)Gpqa(NYU + Cohere + Anthropic (2023))64.4
Lcr(Artificial Analysis)49.0
Scicode(UIUC + Argonne National Lab (2024))37.8
Coding Index(Artificial Analysis)29.9
Terminalbench V2 126.6
Tau Banking14.4
Intelligence Index(Artificial Analysis)12.7
Hle(Center for AI Safety + Scale AI (2025))11.2
Terminalbench V4 00.0
LLM Stats 分類評分
(LLM Stats (zeroeval))Long Context90
Instruction Following80
Language80
Legal80
Finance80
Healthcare80
Math70
Creativity70
Writing70
Reasoning60
Frontend Development60
General60
Physics50
Agents50
Biology50
Chemistry50
Tool Calling50
Knowledge40
Code40
定價
輸入價格$0.16 / 1M tokens
輸出價格$0.65 / 1M tokens
混合價格(3:1)$0.282 / 1M tokens
速度
Tokens/秒38.9
首Token延遲0.38s
首回答延遲51.85s
供應商價格排行
供應商價格排行
2 個供應商
最便宜: DeepInfra最貴: IBM
供應商輸入輸出
1DeepInfra最便宜
$0
$0
2IBM主要
$0.16
$0.65
比較該模型在不同 API 供應商之間的定價。