跳转到主要内容

Muse Glimmer-30B

Meta开源权重Apache 2.0 · 商用许可

描述

Muse Glimmer-30B is Meta Superintelligence Labs' open-weight, dense multimodal model for autonomous agentic work on consumer hardware. Its 29.6-billion-parameter causal transformer includes a roughly 1.8B-parameter ViT-G/14 perception encoder, accepts interleaved text and images, produces text, and supports a 131,072-token context window. It combines controllable multi-step reasoning, schema-based tool use, coding, failure recovery, and compatibility with agent scaffolds such as OpenClaw and Hermes Agent. Meta also releases 4-bit variants sized for 24 GB and 32 GB systems and a DFlash speculative-decoding drafter. The model is trained on more than 100 languages and released under Apache 2.0.

发布日期
2026-08-10
参数规模
29.6B
上下文长度
131K
支持模态
image, text

能力雷达图

37
general
60
coding
60
reasoning
51
science估算
70
agents
70
multimodal

缺少专门科学评测时,Science 由 LLM Stats 科学得分或推理能力估算。

排行榜排名

领域#排名分数来源
智能体能力模型榜108
41.0
LS
多模态榜91
47.0
LS

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

Siren AgentDojo Utility94.2%自报
CI Memories Coverage64.8%自报
WildClawBench47.6%自报
Siren AgentDojo Attack Success Rate28.4%自报
CI Memories Violation Rate26.4%自报
Tau3 Banking23.5%自报

Chat

Tau2 Airline78.0%自报

Code

PinchBench90.5%自报
SkillsBench44.3%自报

Communication

Tau2 Telecom94.1%自报
Tau2 Retail87.1%自报

General

GDPval-AA953.00 / 3000自报
GAIA243.3%自报

Instruction Following

IFBench77.0%自报

Language

MMLU-ProX80.1%自报

Math

AIME 202694.7%自报

Multimodal

OSWorld-Verified65.9%自报

Reasoning

GPQANYU + Cohere + Anthropic (2023)83.5%自报
AA-LCR80.0%自报
CharXiv-R78.8%自报
SWE-Bench Verified76.0%自报
MCP Atlas75.5%自报
ARC-AGI66.9%自报
Beam 128K65.1%自报
Terminal-Bench 2.151.7%自报
SWE-Bench ProPrinceton NLP (2024)51.2%自报
SciCode43.6%自报
Humanity's Last Exam22.0%自报

Search

DeepSearchQA74.6%自报

Vision

OmniDocBench 1.575.8%自报
ScreenSpot Pro75.4%自报
MMMU-Pro74.0%自报

AA 评测指数

(Artificial Analysis)

暂无 AA 评测数据

LLM Stats 分类评分

(LLM Stats (zeroeval))
Legal
100
Finance
100
Agents
96
Reasoning
48
General
37
Communication
90
Chat
80
Instruction Following
80
Language
80
Long Context
80
Structured Output
80
Frontend Development
80
Grounding
80
Healthcare
80
Multimodal
70
Search
70
Spatial Reasoning
70
Tool Calling
70
Vision
70
Math
60
Physics
60
Biology
60
Chemistry
60
Code
60
Coding
50

定价

输入价格$0.3 / 1M tokens
输出价格$1.2 / 1M tokens
混合价格(3:1)$0.525 / 1M tokens
缓存读取价格$0.04 / 1M tokens

速度

暂无速度数据

供应商价格排行

供应商价格排行

10 个供应商

最便宜: DeepInfra最贵: AI/ML API
供应商输入输出
1DeepInfra最便宜
$0
$0
2EmpirioLabs AI
$0.2
$0.8
3Vultr
$0.25
$1
4Meta主要
$0.3
$1.2
5Kilo Gateway
$0.3
$1.2
6DevPass (LLM Gateway)
$0.3
$1.2
7NanoGPT
$0.35
$1.5
8OpenRouter
$0.35
$1.5
9Vercel AI Gateway
$0.35
$1.5
10AI/ML API
$0.3703
$1.587

比较该模型在不同 API 供应商之间的定价。

外部链接