MiMo-V2.5-Pro-Base

2026-04-27

Date the weights (or, if earlier, the technical report) were first made public, YYYY-MM-DD, as dated by the publisher. A Hugging Face repository creation date is used only when no dated publisher page exists, with a note.

mit

License of the released weights as declared by the publisher (model card metadata or license file), e.g. apache-2.0, mit, or the name of a custom license.

text

Inferred from the source; see the explanation below.

Input modalities of the released checkpoint as the publisher states them, from: text, image, video, audio, joined with ' + ' in that order ("text", "text + image + video").

1.02T parameters

Total parameter count as reported by the publisher. See params.note for whether embeddings, MTP heads or vision towers are included.

42B parameters

Parameters used per token in the forward pass, as reported by the publisher.

Not disclosed

Active parameters excluding the input embedding and the output head (unembedding). Stated by the publisher, or computed from params.active by subtracting vocab × d_model for each of the embedding and output head that params.active is stated to include (basis inferred, note with the arithmetic). Not disclosed when the scope of params.active is not stated.

1.021T parameters

Calculated: embedding + output head + all experts + other weights, XiaomiMiMo/MiMo-V2.5-Pro-Base@1e94ccf safetensors headers; reproduces stated total (backbone) and active (with embedding and output head)

Parameters of the text backbone counted from the released checkpoint's safetensors headers (embedding, output head, all experts and other weights; excluding MTP modules, vision/audio towers and n-gram lookup tables). Shown only when the counts reproduce the publisher's stated total and active parameters within 5% under some scope.

1.875B parameters

Calculated: input embedding + output head, XiaomiMiMo/MiMo-V2.5-Pro-Base@1e94ccf safetensors headers

Input embedding plus output head (once if tied), counted from the checkpoint.

1.019T parameters

Calculated: dense + routed backbone weights; excludes input embedding, output head, lookup tables, MTP and other modalities, XiaomiMiMo/MiMo-V2.5-Pro-Base@1e94ccf safetensors headers; reproduces stated total (backbone) and active (with embedding and output head)

Language backbone weights excluding input embeddings, output head, lookup/Engram tables, MTP modules and vision/audio towers. Equals params.counted_total minus params.counted_embedding. Same validation as params.counted_total.

14.77B parameters

Calculated: max(sum of dense and routed weights sharing a full layers.<index> tensor path), all experts included, XiaomiMiMo/MiMo-V2.5-Pro-Base@1e94ccf safetensors headers

Largest group of language-backbone weights sharing a full layers.<index> tensor path, including every routed expert. Excludes lookup/Engram tables, embeddings, output head, MTP and other modalities. Hybrid checkpoints may store mixers and FFNs as separate layers. All-gather volume also depends on wrapping, sharding and dtype.

40.02B parameters

Calculated: non-expert weights + routed experts × 8/384, excluding embedding and output head, XiaomiMiMo/MiMo-V2.5-Pro-Base@1e94ccf safetensors headers; reproduces stated total (backbone) and active (with embedding and output head)

Active parameters per token excluding the input embedding and output head, counted from the checkpoint as all non-expert weights plus routed-expert weights × top_k / experts. Same validation as params.counted_total.

Publisher's rounded figures; whether the three MTP modules are included is not stated.

What the publisher's counts include or exclude (MTP modules, embeddings, vision encoder, n-gram tables).

70

Number of transformer blocks in the main backbone (attention or token mixer plus FFN), excluding MTP modules.

6,144

Residual stream width.

152,576

Embedding table rows (vocab_size in config, may include padding).

No

Input embedding and output head share weights.

87.7714

Calculated: d_model / layers

Width-to-depth ratio, d_model divided by the number of blocks.

GQA, SWA

Token-mixing layer types in the backbone, e.g. GQA, MLA, SWA, Gated DeltaNet, Mamba-2, DSA (DeepSeek sparse attention).

Global at layers 0, 7, 15, 23, 31, 39, 47, 55, 62, 69; SWA elsewhere (0-based)

Inferred from the source; see the explanation below.

Token-mixer order, summarized as a pattern or layer positions. The fact's optional sequence lists each mixer in order; a ratio alone does not establish the order. Auxiliary prediction modules and standalone FFNs are excluded.

10

Layers whose KV cache grows with full context (dense or sparse global attention, MLA). Excludes sliding-window and linear/recurrent layers.

6

Calculated: attn.local_layers / attn.global_layers

attn.local_layers divided by attn.global_layers, the sliding-window to global interleave ratio (5 for '5 SWA : 1 global').

60

Layers with a fixed attention window.

128

Query heads in the global-attention layers.

8

Key/value heads in the global-attention layers. Not applicable to MLA, whose cache is a shared latent.

24,576

Calculated: heads × head_dim

attn.heads × attn.head_dim, the query projection width of global layers.

16

Calculated: heads / kv_heads

attn.heads divided by attn.kv_heads in global layers.

192

Query/key dimension per head in global-attention layers (for MLA, nope + rope dims).

128

Value dimension per head in global layers (MLA v_head_dim; stated for MLA even when equal to the QK head dim).

128 tokens

Sliding-window size of the local layers.

8

KV heads in sliding-window layers (stated even when equal to attn.kv_heads).

128

Query heads in sliding-window layers (stated even when equal to attn.heads).

192

Query/key head dimension in sliding-window layers. Value head dimensions may differ; see attn.local_v_head_dim.

128

Value head dimension in sliding-window layers, when separately specified. KV formulas use this value when present, otherwise the explicit global value dimension, or the local QK dimension when no separate value dimension is given.

partial RoPE (64 of 192 QK dims; base 1e7 global, 1e4 SWA)

Positional scheme in global layers: RoPE, partial RoPE, NoPE, or mixed per layer type.

10,000,000

RoPE base frequency of the global layers at release.

10,000

RoPE base frequency of the sliding-window layers (stated even when equal to attn.rope_theta).

0.334

Fraction of each head's QK dimensions that are rotated.

No

Queries and keys are normalized per head before the dot product.

values × 0.612 in every attention layer; QK scale 1/sqrt(192)

Query-key scaling when it is not 1/sqrt(head_dim), e.g. query_pre_attn_scalar, length-dependent (SSMax-style) scaling, or value scaling.

learnable per-head sink bias in SWA layers; none in global layers

Learned sink logits or sink bias in the softmax, with the layer types that use them.

No

Query/key/value projections have bias terms.

SwiGLU

FFN activation (SwiGLU, GeGLU, ...).

16,384

Intermediate width of the dense FFN layers that are actually executed.

No

FFN projections have bias terms.

2.66667

Calculated: ffn.dim / d_model

ffn.dim divided by shape.d_model.

384

Routed experts per MoE layer.

8

Routed experts selected per token.

0

Always-on experts per MoE layer.

2,048

Intermediate width of one routed expert.

1

Number of initial blocks that use a dense FFN instead of MoE.

sigmoid

Gating function over expert logits (softmax, sigmoid).

Not disclosed

Load-balancing method, with coefficients where disclosed.

Yes

Selected expert weights are renormalized to sum to 1.

0.333333

Calculated: moe.expert_dim / d_model

moe.expert_dim divided by shape.d_model.

48

Calculated: experts / top_k

Routed experts divided by experts selected per token.

0.0411765

Calculated: params.active / params.total

params.active divided by params.total.

Not disclosed

Additional future tokens predicted during pretraining (0 when the model was trained without MTP).

dense

FFN in the MTP module: 'dense' or 'MoE', with its width when it differs from the backbone.

Yes

The released checkpoint contains the MTP module weights.

RMSNorm

Normalization layer type in the blocks.

pre

Pre-norm, post-norm, or sandwich (pre and post).

1e-05

Epsilon of the block normalization layers.

32.77K tokens

Sequence length of the main pretraining stage.

Not disclosed

Longest sequence length trained on after context extension.

262.1K tokens

Longest context the publisher says the model supports, possibly beyond trained length via RoPE scaling.

extended from 32K up to 1M (stages not stated)

How context was extended (stages, RoPE base change, YaRN factor).

27T tokens

Tokens seen in pretraining including annealing and context extension, as reported.

Not disclosed

Optimizer used for the bulk of parameters.

Not disclosed

Peak learning rate of the main pretraining stage.

Not disclosed

Learning rate at the end of the decay.

Not disclosed

Learning-rate schedule shape with phase boundaries.

Not disclosed

Warmup length in steps or tokens.

Not disclosed

Global batch size, including any batch-size ramp: sequences × sequence length (token total), or the token total alone when only that is stated.

Not disclosed

Decoupled weight decay coefficient.

Not disclosed

(beta1, beta2) of the Adam-family optimizer or its AdamW fallback.

Not disclosed

Global gradient-norm clipping threshold.

FP8 mixed precision

Matmul precision of pretraining (BF16, FP8 recipe, NVFP4 recipe) with the scaling granularity where disclosed. Never inferred from the released checkpoint dtype.

26.4706

Calculated: train.tokens / params.total

train.tokens divided by params.total.

674.662

Calculated: train.tokens / params.counted_active_non_embedding

train.tokens divided by params.counted_active_non_embedding, or by the stated params.active_non_embedding when there is no validated checkpoint count.

51,500 bytes

Calculated: kv.bytes_at_context / 131072 tokens; uniform BF16, payload only; quantization overhead excluded. Global KV: 10 × (2560 × 16/8) B/entry; SWA KV: 60 × (2560 × 16/8) B/entry

Retained KV cache divided by context length, including bounded SWA caches and indexer keys. Exports use uniform BF16 at 131,072 tokens. The website can select uniform payload precision (no quantization overhead) or sourced component formats (including scales and documented padding). This is cache payload, not peak prefill memory or persistent prefix storage.

51,200 bytes

Calculated: 51200 B / 1 tokens (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 10 × (2560 × 16/8) B/entry; SWA KV: 60 × (2560 × 16/8) B/entry

Long-context global KV bytes per token, including indexer keys; excludes bounded SWA caches. Compression rates and cross-layer cache reuse are included. Exports use uniform BF16; the website's storage selector also applies here. This is the field to compare with global-KV figures in reports, using the matching storage recipe.

1,226 bytes

Calculated: kv.bytes_per_token (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 10 × (2560 × 16/8) B/entry; SWA KV: 60 × (2560 × 16/8) B/entry

kv.bytes_per_token divided by active parameters in billions (131,072 tokens in exports).

160,719,238 bytes

Calculated: kv.bytes_at_context (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 10 × (2560 × 16/8) B/entry; SWA KV: 60 × (2560 × 16/8) B/entry

kv.bytes_at_context divided by active parameters in billions (131,072 tokens in exports).

6,750,208,000 bytes

Calculated: 51200 B × 131072 (Global KV) + 307200 B × min(131072, 128) (SWA KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 10 × (2560 × 16/8) B/entry; SWA KV: 60 × (2560 × 16/8) B/entry

Retained KV cache for one sequence at the selected context, including SWA windows and indexer keys. Exports use uniform BF16 at 131,072 tokens. Excludes recurrent state, compressor tail state, draft/MTP caches, temporary prefill tensors, allocator metadata and persistent prefix snapshots. Does not estimate peak prefill or total serving memory.

26

Artificial Analysis Intelligence Index v4.3.2 of the post-trained model, from a dated Wayback snapshot of its Artificial Analysis page. The note names the variant (reasoning setting).