Latest

Every published item, in order · 6 items

Blogs News Posts Papers Papers →

10月21日

Models US vs ChinaMulti-source74°13:20

Chinese content美国开源权重模型一周连发三款：Nemotron 3.5、Muse Spark Open、Beam 1.0

美国开源模型一周发了三款，Meta 的 Muse Spark Open 拿下 38.9 分刷新纪录，NVIDIA 和 Reflection 的新模型也在追赶，智能与参数效率对比值得看看。

#Nemotron #Muse Spark Open #Beam 1.0 #开源模型

Event

官方账号Artificial Analysis@ArtificialAnlys11 sourcesSource

10月20日

Models US vs ChinaSelected01:18

Reflection AI's Beam 1.0 Scores 23.5% on ARC-AGI-2 Verified

Cost per task is the notable figure: high reasoning trades just 3.5 points on ARC-AGI-2 for 43% savings, which matters for teams budgeting reasoning workloads. Researchers comparing frontier abstract-reasoning scores against models like GPT-5 or Gemini can pull the verified ARC Prize entry directly. The ARC-AGI-1 versus ARC-AGI-2 gap also offers a data point on how quickly reasoning models are closing the abstraction gap.

#Reflection AI #Beam 1.0 #ARC-AGI #reasoning models

官方账号ARC Prize@arcprizeSource

10月17日

Models Builders13:35

Chinese contentRun Beam 1.0 501B 本地部署，Atomic Chat 发布 4 档量化版本

Reflection AI 的 501B 模型现在能本地跑了，四档 GGUF 量化从 197GB 压到 134GB，最高那档精度损失不到 6%，有硬件的可以试试。

#Beam 1.0 #Reflection AI #GGUF #量化

@atomic_chat_hq@atomic_chat_hqSource

10月16日

Models Builders23:41

Chinese contentReflection AI 发布 Beam 1.0

Reflection AI 发了 Beam 1.0，官方博客里有完整介绍，想了解细节的直接去博客看。

#Reflection AI #Beam 1.0 #开源模型

lmarena.ai@lmarena_aiSource

Models US vs China22:48

Chinese contentReflection开源Beam 1.0：501B总参数、23B激活

Reflection把Beam 1.0开源了，501B总参但只激活23B，跑起来不算重，不过评分比GLM 5.3 Flash低一些，可以拿来对比试试。

#Reflection #Beam 1.0 #GLM #开源模型

Gorden Sun@Gorden_SunSource

Models US vs ChinaSelected22:06

Reflection AI's Beam-1.0-501B-A23B Ranks #12 in Agent Arena

24 of the top 25 open models on Code Arena: WebDev come from Chinese labs, and Beam 1.0 is the lone US exception — a notable shift in the open-weights competitive landscape. Builders evaluating Apache 2.0 agentic models can compare Beam directly against Qwen3.8, Kimi K2.5 Thinking and GLM-5 once the full weights drop this month. Its 23B active parameter design targets reasoning efficiency for agent workloads.

#Reflection AI #Beam 1.0 #open-source models #agents

lmarena.ai@lmarena_aiSource