Solar-Open2-250B

2026-07-22

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.

other (Upstage Solar License)

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").

250.3B parameters

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

15B 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.

250.3B parameters

Calculated: embedding + output head + all experts + other weights, upstage/Solar-Open2-250B@9190fbe 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.

1.611B parameters

Calculated: input embedding + output head, upstage/Solar-Open2-250B@9190fbe safetensors headers

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

248.7B parameters

Calculated: dense + routed backbone weights; excludes input embedding, output head, lookup tables, MTP and other modalities, upstage/Solar-Open2-250B@9190fbe 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.

5.188B parameters

Calculated: max(sum of dense and routed weights sharing a full layers.<index> tensor path), all experts included, upstage/Solar-Open2-250B@9190fbe 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.

13.13B parameters

Calculated: non-expert weights + routed experts × 8/320, excluding embedding and output head, upstage/Solar-Open2-250B@9190fbe 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.

48

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

4,096

Residual stream width.

196,608

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

No

Input embedding and output head share weights.

byte-level BPE, 196,608 vocab, digit splitting; inherited unchanged from Solar Open 1

Tokenizer type and lineage as described by the publisher (e.g. byte-level BPE shared with an earlier model).

85.3333

Calculated: d_model / layers

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

GQA, KDA

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

3 KDA : 1 GQA (GQA first in each block of 4)

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.

12

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

3

Calculated: attn.linear_layers / attn.global_layers

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

36

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

64

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.

8,192

Calculated: heads × head_dim

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

8

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).

NoPE

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

No

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

elementwise sigmoid gate on SDPA output (g_proj from layer input), GQA layers

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

No

Query/key/value projections have bias terms.

KDA with negative eigenvalues (β = 2σ(·) ∈ (0, 2)); 64 heads × 128, no KV grouping; short conv 4 on q/k/v; low-rank (rank 128) decay and output-gate projections

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

64

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, ...).

No

FFN projections have bias terms.

320

Routed experts per MoE layer.

8

Routed experts selected per token.

1

Always-on experts per MoE layer.

1,280

Intermediate width of one routed expert.

1,280

Total intermediate width of the shared expert(s).

0

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.

1

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

Yes

Selected expert weights are renormalized to sum to 1.

0.3125

Calculated: moe.expert_dim / d_model

moe.expert_dim divided by shape.d_model.

40

Calculated: experts / top_k

Routed experts divided by experts selected per token.

0.059931

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).

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.

Not disclosed

Sequence length of the main pretraining stage.

1.049M tokens

Longest sequence length trained on after context extension.

1.049M tokens

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

Stage 4 Length Expansion (865B tokens): continued training to 1M on a higher-quality corpus plus long documents and repo-level code; NoPE

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

~12T

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.

Not disclosed

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

selective weight transfer from Solar Open 1: 5.69B params (2.3%) transferred (embedding/output, norms, GQA q/k/v/o, shared expert; KDA q/o re-mapped from GQA q/o); routed experts, router, KDA k/v, decay/write gates, conv and GQA output gate randomly initialized

Weight initialization (std or scheme) as stated by the publisher; config initializer_range alone is not evidence.

10T tokens curated from a 20T deduplicated pool by quality and rarity; real:synthetic 4:6; math and code each ≥15%; English >80%

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

NVIDIA B200 GPUs

Accelerators used for pretraining, count and duration where stated.

B200

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").

2M GPU hours

Pretraining compute as stated (GPU-hours or FLOPs).

merge of four intermediate checkpoints from Stage 4 (Length Expansion)

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

• general pre-training

  Tokens

  10000000000000

  Source: paper · pp.6–9 §3, §3.3, Figure 8

  General Pre-training is the main stage, training on 10T tokens of a broad-coverage, high-quality data mixture

  Curated 10T mixture.

• intensive pre-training

  Tokens

  1000000000000

  Source: paper · pp.6–9 §3, §3.3, Figure 8

  Stage 3 (Intensive Pre-training) raises the quality threshold and trains again on the best 1T tokens of the Stage-2 data

  Best 1T tokens of the stage-2 data re-filtered at a higher quality threshold.

• length expansion

  Tokens

  865000000000

  Seq len

  1048576

  Source: paper · pp.6–9 §3, §3.3, Figure 8

  and Stage 4 (Length Expansion, 865B).

  Higher-quality base corpus plus long documents, repo-level code, reasoning and agent data; four checkpoints merged.

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.

10B-A1B proxy for architecture ablations; 200B-A15B proxy for the weight-transfer initialization

Smaller models trained to choose the architecture or recipe (sizes, token budgets), as disclosed.

49,152 bytes

Calculated: kv.bytes_at_context / 131072 tokens; uniform BF16, payload only; quantization overhead excluded. Global KV: 12 × (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.

49,152 bytes

Calculated: 49152 B / 1 tokens (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 12 × (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.

3,277 bytes

Calculated: kv.bytes_per_token (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 12 × (2048 × 16/8) B/entry

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

429,496,730 bytes

Calculated: kv.bytes_at_context (131072 tokens) / (params.active / 1e9); uniform BF16, payload only; quantization overhead excluded. Global KV: 12 × (2048 × 16/8) B/entry

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

6,442,450,944 bytes

Calculated: 49152 B × 131072 (Global KV); uniform BF16, payload only; quantization overhead excluded. Global KV: 12 × (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.

25

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).