Skip to content
archdiff

models / compare / glm-5-vs-inkling

GLM-5 vs Inkling

How GLM-5 (Z.ai (Zhipu AI), February 2026) and Inkling (Thinking Machines Lab, July 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+++ glm-5@2026-02+0~3−5

GLM-5

Z.ai (Zhipu AI), February 2026, 744B total, 40B 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-normRMSNorm on MLA latents (no per-head QK-Norm) — Residual & normRMSNorm on MLA latents (no per-head QK…Residual & 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 layersLatent attention (MLA) — global · GQA in many modelsLatent attention (MLA)global · GQA in many modelsSparse / 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-headINT4 QAT in SFT, FP8 rollouts — Training & inferenceINT4 QAT in SFT, FP8 rolloutsTraining & 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 used

--- frontier-llm@2026-11+++ inkling@2026-07+0~5−3

Inkling

Thinking Machines Lab, July 2026, 975B total, 41B 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 pre-norm residual stream — Residual & normSingle pre-norm residual streamResidual & normRMSNorm — pre-normRMSNormpre-normQK-Norm — per-head RMSNorm on q, kQK-Normper-head RMSNorm on q, kLearned relative-position bias (no RoPE) — AttentionLearned relative-position bias (no RoP…AttentionGated DeltaNet / KDA — linear · ~3 of every 4 layersGated DeltaNet / KDAlinear · ~3 of every 4 layersGlobal GQA with log-length scaling — AttentionGlobal GQA with log-length s…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 windowShort causal convolutions (SConv) — AttentionShort causal convolutions (SConv)Attention⊕ residual⊕residualRMSNorm — pre-normRMSNormpre-normSigmoid router with shared-expert sink — Mixture of expertsSigmoid router with shared-expert sinkMixture of expertsRouted 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

GLM-5 and Inkling differ on 11 of 27 components.

Blueprint componentGLM-5Inkling

Not used

Same as the blueprintEncoder-free. Images and video are cut into 40 × 40-pixel patches, 2 frames deep. A 4-layer hierarchical MLP (three linear + RMSNorm + GELU stages that fold space and time into channels, then a linear projection) maps each patch to one 6,144-dim token. 16 kHz audio becomes dMel frames: 80 mel bins, each quantized to 16 levels, embedded by lookup and summed per frame.

Single residual streamOne d = 6,144 residual stream with pre-norm sublayers; no hyper-connections in GLM-5 through GLM-5.3.

Single pre-norm residual streamOne d = 6,144 residual stream through 66 pre-norm blocks; no hyper-connections, branch outputs pass a short conv.

RMSNorm on MLA latents (no per-head QK-Norm)Queries and keys are normalized as compressed latents (2,048 and 512 dims) before up-projection, not per head.

Same as the blueprintPer-head RMSNorm on q and k (head dim 128) in all 66 layers, applied to keys after their short convolution. Because q and k are normalized, the softmax scale is 1/d_h = 1/128 instead of 1/√d_h.

Same as the blueprintDecoupled partial RoPE as in DeepSeek MLA: 64 of each head's 256 q/k dims rotate, using one 64-dim rotary key shared by all heads, with interleaved pairs and base θ = 1,000,000. The DSA indexer rotates 64 of its 128 dims. GLM-5.2 raises θ to 8,000,000 for its 1M window; neither config uses YaRN.

Learned relative-position bias (no RoPE)Each head adds a bias by backward distance: a query-dependent mix of 16 learned profiles, over 512 (local) or 1,024 (global) tokens.

Same as the blueprintMLA in all 78 layers and the MTP layer, with 64 heads: 2,048-dim query latent, 512-dim KV latent plus a 64-dim rotary key (576 cached values per token per layer), q/k head dim 256 (192 + 64 rotary), v head dim 256. This is the MLA-256 variant: head dim raised from 192 to 256 with one-third fewer heads, keeping training FLOPs and parameters roughly constant while lowering decode compute.

Global GQA with log-length scaling11 full-attention layers (every 6th, incl. the last): 64 Q / 8 KV heads, d_h = 128; logits scaled up logarithmically past 128K.

Same as the blueprintDSA in every layer, not alternated with MLA: a lightning indexer (32 heads × 128 dims, ReLU scores with learned per-head weights) selects the top 2,048 tokens per query and MLA attends only to those. It was added after mid-training: 1,000 warm-up steps on 14 × 202,752-token sequences, then 20B tokens of sparse adaptation. The indexer is frozen in RL and uses a deterministic torch.topk. GLM-5.2 runs it in 21 of 78 layers.

Not used

Not used

Same as the blueprint55 of 66 layers (all except 5, 11, …, 65) attend to the last 512 tokens with 64 query heads and 16 KV heads of dimension 128. Keys and values pass through a kernel-4 causal convolution before caching, and the relative-position bias spans the whole window.

Not used

Short causal convolutions (SConv)Depthwise causal convs (kernel 4, with skip) on K, V and on both sublayer outputs before they rejoin the residual.

Same as the blueprintSigmoid affinities over 256 experts with a per-expert correction bias that affects only the top-8 selection (noaux_tc, no group-limited routing). The selected scores are renormalized to sum to 1 and multiplied by a routed scaling factor of 2.5. GLM-5.2's config pins the router to FP32.

Sigmoid router with shared-expert sinkScores 256 routed + 2 shared experts; bias-steered top-6; weights normalized jointly over the 8 active experts, then ×8.

INT4 QAT in SFT, FP8 rolloutsINT4 quantization-aware training during SFT with a bitwise-identical kernel; FP8 inference for RL rollouts and release.

Same as the blueprintTraining precision is not published. Weights ship in BF16 and as an NVFP4 checkpoint (W4A4, 16-element blocks with E4M3 scales) that quantizes only the routed experts of MoE layers 3–65. Attention, shared experts, routers, dense MLPs, the first MoE layer's experts, embeddings, head, norms and convolutions stay BF16, and the KV cache is not quantized. Whether NVFP4 was trained in (QAT) is not stated.

Same as the blueprintThe last post-training stage is on-policy cross-stage distillation: final checkpoints of earlier SFT and RL stages are teachers, and the advantage is replaced by the stop-gradient log-ratio log π_teacher − log π_train, with group size 1 and batch 1,024. GLM-5.2 merged more than ten expert models this way in about two days.

Not disclosed

More comparisons