Ling-3.0-flash-base

2026-08-12

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").

124B parameters

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

5.1B parameters

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

5.1B parameters

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.

124.4B parameters

Calculated: embedding + output head + all experts + other weights, inclusionAI/Ling-3.0-flash-base@0a96b0a safetensors headers; reproduces stated total (backbone) and active (with 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.

804.8M parameters

Calculated: input embedding + output head, inclusionAI/Ling-3.0-flash-base@0a96b0a safetensors headers

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

123.6B parameters

Calculated: dense + routed backbone weights; excludes input embedding, output head, lookup tables, MTP and other modalities, inclusionAI/Ling-3.0-flash-base@0a96b0a safetensors headers; reproduces stated total (backbone) and active (with 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.

3.09B parameters

Calculated: max(sum of dense and routed weights sharing a full layers.<index> tensor path), all experts included, inclusionAI/Ling-3.0-flash-base@0a96b0a 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.

4.701B parameters

Calculated: non-expert weights + routed experts × 8/512, excluding embedding and output head, inclusionAI/Ling-3.0-flash-base@0a96b0a safetensors headers; reproduces stated total (backbone) and active (with 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.

Rounded publisher figures; active count excludes embeddings. Whether the MTP layer is counted is not stated; the checkpoint index (254.97 GB in BF16, ≈127.5B parameters) includes it.

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

42

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

2,560

Residual stream width.

157,184

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

No

Input embedding and output head share weights.

60.9524

Calculated: d_model / layers

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

KDA, MLA

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

5 KDA : 1 MLA

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.

7

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

5

Calculated: attn.linear_layers / attn.global_layers

attn.linear_layers divided by attn.global_layers (3 for '3 Gated DeltaNet : 1 GQA').

35

Layers with a fixed-size recurrent state (linear attention, DeltaNet variants, Mamba).

32

Query heads in the global-attention layers.

6,144

Calculated: heads × head_dim

attn.heads × attn.head_dim, the query projection width of 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).

512

Width of the compressed KV latent that MLA caches per token per layer.

0

Width of the low-rank query projection in MLA; 0 when queries use a full-rank projection.

64

Width of the shared positional key that MLA caches next to the latent.

partial RoPE (MLA, decoupled 64 of 192 QK dims, interleaved), NoPE (KDA)

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

6,000,000

RoPE base frequency of the global layers at release.

No

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

head-wise sigmoid on the MLA output, before the output projection

Gate applied to the attention output before the output projection, with its placement.

No

Query/key/value projections have bias terms.

KDA: 32 heads × 128, short conv 4 + SiLU on q/k/v, L2-normed q/k, full-rank decay and output-gate projections, safe gate with lower bound −5, sigmoid-gated RMSNorm on output

Configuration of the linear / SSM layers (e.g. Gated DeltaNet heads and dims, Mamba-2 state size), as one string.

32

Value heads (or SSM heads) in the linear / SSM layers.

128

Head dimension of the linear / SSM layers (value head dim for DeltaNet variants, head_dim for Mamba-2).

4

Kernel size of the causal short convolution in linear / SSM layers or attention projections.

SwiGLU

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

6,144

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

No

FFN projections have bias terms.

2.4

Calculated: ffn.dim / d_model

ffn.dim divided by shape.d_model.

512

Routed experts per MoE layer.

8

Routed experts selected per token.

1

Always-on experts per MoE layer.

768

Intermediate width of one routed expert.

768

Total intermediate width of the shared expert(s).

2

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.

2.5

Multiplier applied to the combined routed-expert output (routed_scaling_factor).

Yes

Selected expert weights are renormalized to sum to 1.

8 groups, top-4 groups per token

Expert groups and groups selected per token, or node-limited routing, when used.

0.3

Calculated: moe.expert_dim / d_model

moe.expert_dim divided by shape.d_model.

64

Calculated: experts / top_k

Routed experts divided by experts selected per token.

0.041129

Calculated: params.active / params.total

params.active divided by params.total.

1

Inferred from the source; see the explanation below.

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

1

MTP modules with distinct weights. 1 with mtp.depth > 1 means one module is applied recursively.

RMSNorm(embedding of token t+1) and RMSNorm(hidden) concatenated, eh_proj 2d→d, then one gated MLA + MoE block with final RMSNorm

Contents of one MTP module (mixer and FFN type, norms, how the previous hidden state and next-token embedding are combined).

MoE

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

Yes

MTP modules use the main model's input embedding and output head.

Not disclosed

Weight of the MTP loss, including its schedule.

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-06

Epsilon of the block normalization layers.

8.192K tokens

Sequence length of the main pretraining stage.

262.1K tokens

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.

8K → 32K → 256K; RoPE base 10,000 in the pretrained-only checkpoint, 6,000,000 after mid-training

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

Not disclosed

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.

Warmup-Stable and Merge (WSM): warmup, stable LR, no decay phase; checkpoint merging replaces decay

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.

>10T text tokens; mainly English and Chinese: Common Crawl, academic papers, developer Q&A and code, licensed text, own crawl (Ant Spider), synthetic data from Ling-2.6 models

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

~2025-10

Pretraining data cutoff date.

WSM: weighted checkpoint merging of mid-training checkpoints replaces LR decay

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

• pretraining

  Seq len

  8192

  Source: model-card · README.md; Introduction

  Pretrained checkpoint have completed large-scale pretraining but have not undergone mid-training, WSM merging (or learning-rate decay), or post-training.

  Released as Ling-3.0-flash-base-30T; seq_len from that checkpoint's config (max_position_embeddings 8192).

• mid-training

  Source: model-card · README.md; Introduction

  Mid-trained checkpoint have completed mid-training but have not undergone WSM merging (or learning-rate decay) or post-training.

  Released as Ling-3.0-flash-base-midtrain; its config has max_position_embeddings 262144 and rope_theta 6,000,000.

• WSM merge

  Source: model-card · README.md; Introduction

  Merged checkpoints have undergone

  Checkpoint merge in place of LR decay; released as Ling-3.0-flash-base.

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.

8,064 bytes

Calculated: kv.bytes_at_context / 131072 tokens; uniform BF16, payload only; quantization overhead excluded. Global KV: 7 × (576 × 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.

8,064 bytes

Calculated: 8064 B / 1 tokens (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 7 × (576 × 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.

1,581 bytes

Calculated: kv.bytes_per_token (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 7 × (576 × 16/8) B/entry

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

207,247,962 bytes

Calculated: kv.bytes_at_context (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 7 × (576 × 16/8) B/entry

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

1,056,964,608 bytes

Calculated: 8064 B × 131072 (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 7 × (576 × 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.

20

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).