Skip to content
archdiff

models / compare / minimax-m2-vs-deepseek-v4

MiniMax-M2.7 vs DeepSeek-V4-Pro

How MiniMax-M2.7 (MiniMax, March 2026) and DeepSeek-V4-Pro (DeepSeek, April 2026) are built, drawn on the same November 2026 blueprint. They make different choices on 11 of 27 components. Select a component in either diagram to find it in the table.

--- frontier-llm@2026-11+++ minimax-m2@2026-03+0~4−7

MiniMax-M2.7

MiniMax, March 2026, 229.9B total, 9.8B active

Text · code · tool outputText · code · tool outputImages · video · audioImages · video · audioByte-level BPE tokenizer — 128K–262K vocabByte-level BPE tokenizer128K–262K vocabVision / audio encoder — native ViT or encoder-free patchesVision / audio encodernative ViT or encoder-free patchesToken embedding — media embeds spliced in placeToken embeddingmedia embeds spliced in placeEngram n-gram memory — hashed 2–4-gram lookup (V4.1)Engram n-gram memoryhashed 2–4-gram lookup (V4.1)Single residual stream — Residual & normSingle residual streamResidual & normRMSNorm — pre-normRMSNormpre-normLayer-wide QK RMSNorm — Residual & normLayer-wide QK RMSNormResidual & normPartial RoPE / NoPE — rotate 64 dims, or nonePartial RoPE / NoPErotate 64 dims, or noneGated DeltaNet / KDA — linear · ~3 of every 4 layersGated DeltaNet / KDAlinear · ~3 of every 4 layersFull grouped-query attention in all 62 layers — AttentionFull grouped-query attention…AttentionSparse / compressed attn — DSA indexer · CSA / HCASparse / compressed attnDSA indexer · CSA / HCASliding-window attention — local · 128–1,024-token windowSliding-window attentionlocal · 128–1,024-token windowOutput gate + sink — σ(x·W_g) ⊙ o · learnable sinkOutput gate + sinkσ(x·W_g) ⊙ o · learnable sink⊕ residual⊕residualRMSNorm — pre-normRMSNormpre-normMoE router — σ / √softplus · bias balancingMoE routerσ / √softplus · bias balancingRouted experts — top-6–22 of 256–896 · SwiGLURouted expertstop-6–22 of 256–896 · SwiGLUShared expert — 1–2 · always activeShared expert1–2 · always active⊕ residual⊕residualFinal RMSNormFinal RMSNormLM head — untied · vocab logitsLM headuntied · vocab logitsMTP module(s) — predict t+2 … t+kMTP module(s)predict t+2 … t+kNext-token logitsNext-token logitsDraft tokensDraft tokensMuon optimizer — Newton–Schulz · QK-Clip · per-headMuon optimizerNewton–Schulz · QK-Clip · per-headFP8 block-scaled checkpoint — Training & inferenceFP8 block-scaled checkpointTraining & inferenceLong-context extension — staged 4K → 1M · YaRNLong-context extensionstaged 4K → 1M · YaRNRL with verifiable rewards — GRPO family · async agentic RLRL with verifiable rewardsGRPO family · async agentic RLOn-policy distillation — reverse KL to specialist teachersOn-policy distillationreverse KL to specialist teachersSpeculative decoding — MTP self-drafting · DSparkSpeculative decodingMTP self-drafting · DSparkKV-cache compression — FP4 KV · cross-layer reuse · SSDKV-cache compressionFP4 KV · cross-layer reuse · SSD
AttentionMixture of expertsResidual & normInput & outputTraining & inferenceData flowResidual / skipAddedReplacedNot usedNot disclosed

--- frontier-llm@2026-11+++ deepseek-v4@2026-04+0~1−3

DeepSeek-V4-Pro

DeepSeek, April 2026, 1.6T total, 49B active

Text · code · tool outputText · code · tool outputImages · video · audioImages · video · audioByte-level BPE tokenizer — 128K–262K vocabByte-level BPE tokenizer128K–262K vocabVision / audio encoder — native ViT or encoder-free patchesVision / audio encodernative ViT or encoder-free patchesToken embedding — media embeds spliced in placeToken embeddingmedia embeds spliced in placeEngram n-gram memory — hashed 2–4-gram lookup (V4.1)Engram n-gram memoryhashed 2–4-gram lookup (V4.1)Residual streams — mHC: n = 4 mixed streamsResidual streamsmHC: n = 4 mixed streamsRMSNorm — pre-normRMSNormpre-normQK-Norm — per-head RMSNorm on q, kQK-Normper-head RMSNorm on q, kPartial RoPE / NoPE — rotate 64 dims, or nonePartial RoPE / NoPErotate 64 dims, or noneGated DeltaNet / KDA — linear · ~3 of every 4 layersGated DeltaNet / KDAlinear · ~3 of every 4 layersHeavily Compressed Attention (HCA) — AttentionHeavily Compressed Attention…AttentionSparse / compressed attn — DSA indexer · CSA / HCASparse / compressed attnDSA indexer · CSA / HCASliding-window attention — local · 128–1,024-token windowSliding-window attentionlocal · 128–1,024-token windowOutput gate + sink — σ(x·W_g) ⊙ o · learnable sinkOutput gate + sinkσ(x·W_g) ⊙ o · learnable sink⊕ residual⊕residualRMSNorm — pre-normRMSNormpre-normMoE router — σ / √softplus · bias balancingMoE routerσ / √softplus · bias balancingRouted experts — top-6–22 of 256–896 · SwiGLURouted expertstop-6–22 of 256–896 · SwiGLUShared expert — 1–2 · always activeShared expert1–2 · always active⊕ residual⊕residualFinal RMSNormFinal RMSNormLM head — untied · vocab logitsLM headuntied · vocab logitsMTP module(s) — predict t+2 … t+kMTP module(s)predict t+2 … t+kNext-token logitsNext-token logitsDraft tokensDraft tokensMuon optimizer — Newton–Schulz · QK-Clip · per-headMuon optimizerNewton–Schulz · QK-Clip · per-headFP8 → FP4 training — block-scaled FP8 · MXFP4/NVFP4 QATFP8 → FP4 trainingblock-scaled FP8 · MXFP4/NVFP4 QATLong-context extension — staged 4K → 1M · YaRNLong-context extensionstaged 4K → 1M · YaRNRL with verifiable rewards — GRPO family · async agentic RLRL with verifiable rewardsGRPO family · async agentic RLOn-policy distillation — reverse KL to specialist teachersOn-policy distillationreverse KL to specialist teachersSpeculative decoding — MTP self-drafting · DSparkSpeculative decodingMTP self-drafting · DSparkKV-cache compression — FP4 KV · cross-layer reuse · SSDKV-cache compressionFP4 KV · cross-layer reuse · SSD
AttentionMixture of expertsResidual & normInput & outputTraining & inferenceData flowResidual / skipAddedReplacedNot usedNot disclosed

Component by component

MiniMax-M2.7 and DeepSeek-V4-Pro differ on 11 of 27 components.

Blueprint componentMiniMax-M2.7DeepSeek-V4-Pro

Single residual streamOne d = 3,072 residual stream with pre-norm sublayers; no hyper-connections.

Same as the blueprintmHC with n = 4 streams of width 7,168 around every attention and MoE sublayer. The residual-mixing matrix is projected to doubly stochastic with 20 Sinkhorn–Knopp iterations; read-in uses σ(·) and write-back 2σ(·). DeepSeek reports a 6.7% wall-time overhead per overlapped 1F1B pipeline stage.

Layer-wide QK RMSNormOne RMSNorm over all query heads (6,144 dims) and one over all key heads (1,024 dims), applied before RoPE.

Same as the blueprintAn RMS normalization of each of the 128 query heads and an RMSNorm on the single 512-dim KV entry, applied just before core attention in both CSA and HCA. DeepSeek credits this for training with Muon without QK-Clip.

Full grouped-query attention in all 62 layersCausal softmax over the whole context in every layer: 48 query heads share 8 KV heads of dim 128; no window, sink or gate.

Heavily Compressed Attention (HCA)Every 128 tokens pooled into one KV entry; dense shared-KV attention over all of them plus a 128-token window.

Not used

Same as the blueprintCompressed Sparse Attention in 30 of 61 layers (even layers 2–60), interleaved with HCA. Each KV entry pools 8 overlapping tokens at stride m = 4 with a per-channel softmax. A lightning indexer (64 heads × 128 dims, ReLU scores weighted per head) keeps the top 1,024 entries, which the 128 query heads attend to together with the SWA entries.

Not used

Same as the blueprintV4-Pro has no pure sliding-window layers (V4-Flash makes its first two layers pure SWA). Every CSA and HCA layer adds 128 uncompressed KV entries for the most recent tokens to the same softmax, which covers the tokens in the query's own not-yet-compressed block.

Not used

Same as the blueprintNo output gate. Each of the 128 heads has a learnable sink logit whose exponential is added to the softmax denominator in both CSA and HCA, so its attention weights can sum to less than 1.

Not used

Same as the blueprintOne always-active shared SwiGLU expert of intermediate size 3,072 in every MoE layer, with the same clamping; unlike the routed experts it is not stored in FP4.

Not disclosed

Same as the blueprintMuon for most matrices: 10 hybrid Newton–Schulz steps (8 with coefficients (3.4445, −4.7750, 2.0315), then 2 with (2, −1.5, 0.5)), Nesterov momentum 0.95, weight decay 0.1, update RMS rescaled to 0.18. AdamW (β = 0.9 / 0.95) trains the embedding, prediction head, RMSNorm gains and mHC static biases and gates. No QK-Clip.

FP8 block-scaled checkpointOpen weights ship as FP8 E4M3 with 128 × 128 block scales; router, its bias, embeddings and LM head stay in higher precision.

Same as the blueprintFP8 mixed-precision training, with most released weights in FP8. In post-training, MXFP4 quantization-aware training covers the routed-expert weights and the CSA indexer's QK path, which is cached and multiplied in FP4; index scores drop to BF16 for a 2× faster top-k at 99.7% recall. FP4 expert weights dequantize losslessly to FP8, so the FP8 training stack is reused.

Not disclosed

Same as the blueprintMore than ten specialist teachers are merged into one student by on-policy distillation with a weighted, full-vocabulary reverse-KL loss. Teacher logits are rebuilt on the fly from cached last-layer hidden states instead of being stored.

Same as the blueprintThe 3 MTP modules draft tokens that the main model verifies in one forward pass. During RL they are co-trained with the policy via a top-K KL loss so acceptance stays high as the policy shifts.

Not disclosed

More comparisons