跳转到主要内容

Granite 4.2 3B

IBM开源权重Apache 2.0 · 商用许可

描述

IBM Granite 4.2 3B is the compact 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. Unlike the 8B and 30B, it did not receive the extra agentic RL block. Open weights on Hugging Face under Apache 2.0. Text-only (not multimodal).

发布日期
2026-08-25
参数规模
3.0B
上下文长度
—
支持模态
text

能力雷达图

9
general
19
coding
56
reasoning
35
science
50
agents
0
multimodal

排行榜排名

领域#排名分数来源
智能体能力模型榜112
35.0
LS
代码能力榜389
23.0
AA
通用能力榜543
11.0
AA
科学能力364
31.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

TAU3-Bench45.8%自报

General

Arena-Hard v235.0%自报

Instruction Following

IFBench74.3%自报

Language

MMLU-Pro67.8%自报

Long Context

RULER 64k67.5%自报
RULER 128k55.3%自报

Math

AIME 202578.3%自报
HMMT2566.7%自报

Reasoning

LiveCodeBench v669.7%自报
GPQANYU + Cohere + Anthropic (2023)54.8%自报
ProfBench32.1%自报
SciCode24.1%自报

Tool Calling

BFCL-V452.4%自报

AA 评测指数

(Artificial Analysis)
Gpqa(NYU + Cohere + Anthropic (2023))
55.9
Scicode(UIUC + Argonne National Lab (2024))
25.3
Lcr(Artificial Analysis)
24.3
Coding Index(Artificial Analysis)
17.5
Terminalbench V2 1
13.9
Intelligence Index(Artificial Analysis)
9.1
Hle(Center for AI Safety + Scale AI (2025))
6.6
Tau Banking
5.6
Terminalbench V4 0
0.0

LLM Stats 分类评分

(LLM Stats (zeroeval))
Instruction Following
70
Language
70
Legal
70
Finance
70
Healthcare
70
Long Context
60
Math
60
General
60
Reasoning
50
Tool Calling
50
Physics
40
Agents
40
Biology
40
Chemistry
40
Knowledge
30
Creativity
30
Writing
30
Code
20

定价

输入价格$0.03 / 1M tokens
输出价格$0.12 / 1M tokens
混合价格(3:1)$0.052 / 1M tokens

速度

Tokens/秒39.4
首Token延迟0.36s
首回答延迟51.17s

供应商价格排行

供应商价格排行

2 个供应商

最便宜: DeepInfra最贵: IBM
供应商输入输出
1DeepInfra最便宜
$0
$0
2IBM主要
$0.03
$0.12

比较该模型在不同 API 供应商之间的定价。

外部链接