frames·cloud

MiMo-V2.5

xiaomi-mimo-v2-5

fields

nameREQ MiMo-V2.5 huggingface.co 2026-05-03
vendorREQ Xiaomi huggingface.co 2026-05-03
released_at 2026-04-27 huggingface.co 2026-05-03
modality text, image, video, audio, code huggingface.co 2026-05-03
access open-weights huggingface.co 2026-05-03
context_window 1000000 huggingface.co 2026-05-03
parameters 310B total / 15B active huggingface.co 2026-05-03
benchmark_score 65.8 huggingface.co 2026-05-03
benchmark_name Terminal-Bench 2.0 huggingface.co 2026-05-03
announcement_url https://huggingface.co/XiaomiMiMo/MiMo-V2.5 huggingface.co 2026-05-03

history · 10 fields · 10 revisions

name1 revision
MiMo-V2.5 current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
vendor1 revision
Xiaomi current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
released_at1 revision
2026-04-27 current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
modality1 revision
text, image, video, audio, code current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
access1 revision
open-weights current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
context_window1 revision
1000000 current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
parameters1 revision
310B total / 15B active current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
benchmark_score1 revision
65.8 current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
benchmark_name1 revision
Terminal-Bench 2.0 current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.
announcement_url1 revision
https://huggingface.co/XiaomiMiMo/MiMo-V2.5 current huggingface.co · 2026-05-03
MiMo-V2.5 is a native omnimodal model with strong agentic capabilities. Architecture: Sparse MoE, 310B total / 15B activated parameters. Context Length: Up to 1M tokens. Modalities: Text, Image, Video, Audio. 21,407 downloads. Terminal-Bench 2.0: 65.8, SWE Bench Pro: 56.1.