Mixture of Experts AI tools: compare listings

All Mixture of Experts tools (19)

DeepSeek V3Previous DeepSeek V3 snapshot. Not a GPT-4o / Claude 3.5 census.
DeepSeek-V4.1-FlashDeepSeek-V4.1-Flash is a MIT 552B multimodal MoE with 8B/16B activation, 1M context, and about 1/4 the KV cache of V4-Flash.
DeepSeek-V4-Flash-Vision-ExpDeepSeek-V4-Flash-Vision-Exp is the first V4 multimodal experiment: MIT vision weights on Flash, 305B params, screenshots plus text agents.
DeepSeek V4 FlashDeepSeek small agent model: 284B/13B MoE, 1M context, MIT weights, $0.14/$0.28 per 1M tokens.
DeepSeek V4DeepSeek V4 represents the next generation of DeepSeek's flagship AI models, building upon the success of V3 with enhanced capabilities in reasoning, multimodal processing, and agent-based interactions.
Gemma 4 26B A4BGoogle DeepMind's open-weight 25.2B MoE model with 3.8B active parameters, 256K context, multimodal input, and a commercially permissive Apache 2.0 license.
GLM-5.3-FlashZhipu's first natively multimodal GLM (confirmed as the 'Ox Alpha' stealth model): 320B MoE with 18B active, 1M context, MIT open weights at $0.15/M input.
Hy4 previewTencent's next-generation open-weight MoE flagship: 770B total params, 49B activated per token, 1M-token context. Apache 2.0, Gated DSA attention, and blind-eval scores edging out GLM 5.3 and Kimi K3.
KAT-Coder V2.5Kwaipilot's open-weight agentic coding model: 35B MoE, 3B active per token, Qwen3.6 base, Apache 2.0, 262K context, top PinchBench tool-use score.
Laguna S 2.1Poolside's open-weight 118B MoE coding model with 8B active parameters, a 1M-token context window, native interleaved reasoning, and an OpenMDW-1.1 license.
Ling-3.0-flash-VLLing-3.0-flash-VL is inclusionAI's MIT multimodal MoE: 124B total, 5.5B active, with image and video.
LongCat 2.0Meituan's open-weight MoE model: 1.6T total / ~48B active params, 1M context, MIT. Strong on coding and agentic tasks (2026-08-25).
MiniMax M2.1Open-source 230B parameter MoE model optimized for multi-language coding, agentic workflows, and real-world development tasks with 74% SWE-bench performance.
NVIDIA Nemotron 3.5 Lightning 30B A3BNVIDIA's efficient open-weight 30B MoE hybrid model with 3B active parameters, 1M-token context, and single-GPU deployment for local reasoning and coding.
Qwen3.8-Flash-NextAlibaba's open-weight architecture preview of Qwen4: 125B multimodal MoE with 6B active plus a 51B N-gram table, 262K native context, at $0.16/M input.
Qwen3.8-MaxAlibaba's 2.4T-parameter open-weight flagship: 95B active MoE, text/image/video input, 1M context, at $2/$6 per million tokens.
Tiel-Coder 35B A3BAgentic coding MoE. Not a lab release.
WeLM-617BWeChat AI's 617B MoE model with 23B active parameters, demonstrating the first sequence-length scaling at frontier scale via Hidden Decoding, powering WeChat.
WizardLM-2 8x22BPrevious WizardLM-2 8x22B snapshot.

Items tagged with Mixture of Experts (19)

DeepSeek V4 Flash

Models

DeepSeek small agent model: 284B/13B MoE, 1M context, MIT weights, $0.14/$0.28 per 1M tokens.

#DeepSeek

45000000DR: 95AS: 88

WeLM-617B

Models

WeChat AI's 617B MoE model with 23B active parameters, demonstrating the first sequence-length scaling at frontier scale via Hidden Decoding, powering WeChat.

#Mixture of Experts

500000DR: 80AS: 75

MiniMax M2.1

Models

Open-source 230B parameter MoE model optimized for multi-language coding, agentic workflows, and real-world development tasks with 74% SWE-bench performance.

#Open Source

500000DR: 68AS: 72

NVIDIA Nemotron 3.5 Lightning 30B A3B

Models

NVIDIA's efficient open-weight 30B MoE hybrid model with 3B active parameters, 1M-token context, and single-GPU deployment for local reasoning and coding.

#NVIDIA

345DR: 100AS: 100

WizardLM-2 8x22B

Models

Previous WizardLM-2 8x22B snapshot.

#Microsoft

100DR: 100AS: 100

DeepSeek V3

Models

Previous DeepSeek V3 snapshot. Not a GPT-4o / Claude 3.5 census.

#DeepSeek

135DR: 95AS: 88

DeepSeek V4 represents the next generation of DeepSeek's flagship AI models, building upon the success of V3 with enhanced capabilities in reasoning, multimodal processing, and agent-based interactions.

#DeepSeek

135DR: 95AS: 88

DeepSeek-V4.1-Flash

Models

DeepSeek-V4.1-Flash is a MIT 552B multimodal MoE with 8B/16B activation, 1M context, and about 1/4 the KV cache of V4-Flash.

#DeepSeek

1000DR: 80AS: 80

DeepSeek-V4-Flash-Vision-Exp

Models

DeepSeek-V4-Flash-Vision-Exp is the first V4 multimodal experiment: MIT vision weights on Flash, 305B params, screenshots plus text agents.

#DeepSeek

1000DR: 80AS: 80

Gemma 4 26B A4B

Models

Google DeepMind's open-weight 25.2B MoE model with 3.8B active parameters, 256K context, multimodal input, and a commercially permissive Apache 2.0 license.

#Google

1000DR: 80AS: 80

GLM-5.3-Flash

Models

Zhipu's first natively multimodal GLM (confirmed as the 'Ox Alpha' stealth model): 320B MoE with 18B active, 1M context, MIT open weights at $0.15/M input.

#Zhipu AI

1000DR: 80AS: 80

Hy4 preview

Models

Tencent's next-generation open-weight MoE flagship: 770B total params, 49B activated per token, 1M-token context. Apache 2.0, Gated DSA attention, and blind-eval scores edging out GLM 5.3 and Kimi K3.

#Tencent

1000DR: 80AS: 80

KAT-Coder V2.5

Models

Kwaipilot's open-weight agentic coding model: 35B MoE, 3B active per token, Qwen3.6 base, Apache 2.0, 262K context, top PinchBench tool-use score.

#Open Weights

1000DR: 80AS: 80

Laguna S 2.1

Models

Poolside's open-weight 118B MoE coding model with 8B active parameters, a 1M-token context window, native interleaved reasoning, and an OpenMDW-1.1 license.

#Open Weights

1000DR: 80AS: 80

Ling-3.0-flash-VL

Models

Ling-3.0-flash-VL is inclusionAI's MIT multimodal MoE: 124B total, 5.5B active, with image and video.

#Open Weights

1000DR: 80AS: 80

LongCat 2.0

Models

Meituan's open-weight MoE model: 1.6T total / ~48B active params, 1M context, MIT. Strong on coding and agentic tasks (2026-08-25).

#Mixture of Experts

1000DR: 80AS: 80

Qwen3.8-Flash-Next

Models

Alibaba's open-weight architecture preview of Qwen4: 125B multimodal MoE with 6B active plus a 51B N-gram table, 262K native context, at $0.16/M input.

#Qwen

1000DR: 80AS: 80