Granite-4.1-30B-Base

2026-04-29

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.

apache-2.0

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

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").

28.87B parameters

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

28.87B parameters

Inferred from the source; see the explanation below.

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.

28.87B parameters

Calculated: embedding + output head + all experts + other weights, ibm-granite/granite-4.1-30b-base@9bb22a9 safetensors headers; reproduces stated total (backbone) and active (with embedding)

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.

411M parameters

Calculated: input embedding + output head, ibm-granite/granite-4.1-30b-base@9bb22a9 safetensors headers

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

28.45B parameters

Calculated: dense + routed backbone weights; excludes input embedding, output head, lookup tables, MTP and other modalities, ibm-granite/granite-4.1-30b-base@9bb22a9 safetensors headers; reproduces stated total (backbone) and active (with embedding)

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.

444.6M parameters

Calculated: max(sum of dense and routed weights sharing a full layers.<index> tensor path), all experts included, ibm-granite/granite-4.1-30b-base@9bb22a9 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.

28.45B parameters

Calculated: all weights except embedding and output head, ibm-granite/granite-4.1-30b-base@9bb22a9 safetensors headers; reproduces stated total (backbone) and active (with embedding)

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.

64

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

4,096

Residual stream width.

100,352

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

Yes

Input embedding and output head share weights.

64

Calculated: d_model / layers

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

GQA

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

64 GQA

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.

64

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

32

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.

4,096

Calculated: heads × head_dim

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

4

Calculated: heads / kv_heads

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

128

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

RoPE

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

50,000,000

RoPE base frequency of the global layers at release.

No

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

0.0078125 (1/head_dim) instead of 1/sqrt(head_dim)

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.

No

Query/key/value projections have bias terms.

SwiGLU

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

32,768

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

No

FFN projections have bias terms.

8

Calculated: ffn.dim / d_model

ffn.dim divided by shape.d_model.

Not disclosed

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

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.

branch outputs × 0.175 before the residual add (residual_multiplier)

Residual design when it is not a plain sum, e.g. hyper-connections with n streams, gated residuals, depth-scaled residuals, FP32 residual accumulation.

× 12 (embedding_multiplier)

Scaling or normalization applied to input embeddings (e.g. × sqrt(d_model), RMSNorm after embedding).

logits ÷ 16 (logits_scaling)

Scaling applied to output logits other than a softcap (e.g. muP-style width multiplier).

4.096K tokens

Sequence length of the main pretraining stage.

524.3K tokens

Longest sequence length trained on after context extension.

524.3K tokens

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

Staged 4K → 32K → 128K → 512K (phase 5, 396B tokens); exponential LR decay from 1e-4 to 0; model merge after each stage

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

15T 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.

0

Learning rate at the end of the decay.

Warmup + power schedule (phases 1-2), exponential decay (phase 3), linear decay to zero (phase 4), exponential decay from 1e-4 to 0 (phase 5, long context)

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.

Not disclosed

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

FIM with prefix and suffix tokens

Objectives beyond next-token prediction and MTP (fill-in-the-middle rate, distillation, multimodal tokens from the start).

Phase 1: CommonCrawl ~59%, code ~20%, technical ~10.5%, math ~7%, multilingual ~2%, domain-specific ~1.5%. Later phases raise math/code and add synthetic, long-CoT and instruction data.

Pretraining data composition as disclosed (domain shares, synthetic fraction, repetition).

NVIDIA GB200 NVL72 cluster (CoreWeave)

Accelerators used for pretraining, count and duration where stated.

GB200

Accelerator count and type for the main pretraining run, as the count with thousands separators and the chip's short name ("2,048 H800", "2,048 GH200"). A stated range or bound stays as stated ("2,048–4,096 GH200").

model merge after each long-context stage

EMA or checkpoint merging applied to produce the released base weights.

• phase-1

  Tokens

  10000000000000

  Source: model-card · README.md; Model Summary and Training Data table

  10 trillion tokens in phase one

  General mixture; warmup and power LR schedule.

• phase-2

  Tokens

  2000000000000

  Source: model-card · README.md; Model Summary and Training Data table

  2 trillion tokens each in phases two and three

  Higher code and math share; power LR schedule.

• phase-3

  Tokens

  2000000000000

  Source: model-card · README.md; Model Summary and Training Data table

  High quality training data, exponential decay of learning rate.

  High-quality data; exponential LR decay.

• phase-4

  Tokens

  500000000000

  Source: model-card · README.md; Model Summary and Training Data table

  0.5 trillion tokens in phase four

  High-quality data; linear LR decay to zero.

• phase-5-long-context

  Tokens

  396000000000

  Source: model-card · README.md; Model Summary and Training Data table

  Long Context Extension with exponential learning rate schedule.

  Sub-stages at 32K, 128K and 512K; exponential LR decay from 1e-4 to 0. Tokens from the card's Training Data table (396B).

Ordered pretraining stages. Each item: name, tokens, seq_len, lr, batch_tokens, batch_sequences, steps, note, quote, and source when it differs from the fact's source.

519.647

Calculated: train.tokens / params.total

train.tokens divided by params.total.

527.154

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.

262,144 bytes

Calculated: kv.bytes_at_context / 131072 tokens; uniform BF16, payload only; quantization overhead excluded. Global KV: 64 × (2048 × 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.

262,144 bytes

Calculated: 262144 B / 1 tokens (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 64 × (2048 × 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.

9,081 bytes

Calculated: kv.bytes_per_token (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 64 × (2048 × 16/8) B/entry

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

1,190,329,853 bytes

Calculated: kv.bytes_at_context (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 64 × (2048 × 16/8) B/entry

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

34,359,738,368 bytes

Calculated: 262144 B × 131072 (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 64 × (2048 × 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.

7

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).