DeepSeek-V4 Architecture: What Actually Changed.

The official 2026 preview: V4-Pro and Flash, CSA+HCA hybrid attention, mHC, Muon, 32T+ tokens, reasoning modes, and million-token efficiency.

This chapter follows the series' four-layer pyramid: intuition first, then consequences, system design, and finally implementation-level checks. It is written to be useful both as a first explanation and as a review sheet before reading the primary papers.

What you should be able to do after reading

• Explain the mechanism without relying on the feature name.
• Trace the relevant tensors, losses, or messages through one concrete example.
• Distinguish a paper claim from an inference, implementation choice, or marketing shorthand.
• Design a minimal experiment that could prove the idea wrong.

Where this chapter fits in the ten-phase map

The final phase combines the entire series: compressed and sparse attention for long context, MoE for capacity, low precision for efficiency, reasoning RL for deliberate behavior, and systems controls for agents. Future claims should be traced back to one of these concrete mechanisms.

The dependency is useful when debugging. If the model-level equation is correct but the measured result is poor, walk backward through representation, numerical format, memory layout, routing or communication, and finally the evaluation harness. The first broken contract is usually more actionable than the final benchmark delta.

1. Two sparse model sizes

The official preview includes V4-Pro at 1.6T total/49B active parameters and V4-Flash at 284B total/13B active. Both support one million tokens. These are open-weight MoE checkpoints, so total storage and active arithmetic remain distinct.

2. Hybrid attention replaces one-size-fits-all

V4 combines Compressed Sparse Attention (CSA) with Heavily Compressed Attention (HCA). The architecture allocates expensive interaction selectively while retaining a highly compressed path, reducing long-context compute and cache rather than applying dense attention everywhere.

3. mHC strengthens residual transport

Manifold-Constrained Hyper-Connections generalize the residual stream while constraining mixing for stability. The goal is improved signal and gradient propagation through deep MoE stacks without permitting unconstrained hyper-connections to destabilize optimization.

4. Muon joins the optimization stack

The report lists the Muon optimizer among the major upgrades, alongside architecture changes. V4 pretrains both models on more than 32T tokens, so convergence efficiency and numerical stability are as important as per-token inference.

5. Efficiency claims are relative and scoped

At one-million-token context, V4-Pro reports 27% of V3.2’s single-token inference FLOPs and 10% of its KV cache. These ratios describe a specific long-context setting, not a universal 73% latency reduction across hardware, batch sizes, and prompt lengths.

Engineering lens. For every concept above, identify the tensor, state, metric, or system boundary that makes it observable. Then ask which assumption would make the claim fail. This keeps the chapter testable instead of leaving it as architecture vocabulary.

Worked example

Compare V3.2 and V4-Pro at 1M context using four ledgers: total weights, active parameters, attention FLOPs for one decoded token, and KV cache. Only the final two receive the published 27% and 10% ratios.

Do the arithmetic with small dimensions first. Small examples expose index shifts, hidden assumptions, and missing denominators that disappear inside a billion-parameter headline. Once the hand-worked result is correct, automate it and compare the program output against the same values.

Implementation and measurement plan

Use the official checkpoint config and encoding package. Benchmark non-think, Think High, and Think Max separately; the model card recommends at least 384K context for Think Max. Validate serving support for hybrid cache layouts before claiming 1M operation.

1. State the exact model, checkpoint, hardware, and date behind every numerical claim.
2. Separate algorithmic complexity, theoretical FLOPs, measured latency, memory, and end-to-end cost.
3. Build a small reference implementation before optimizing kernels or distributing it.
4. Compare against an equal-compute or equal-parameter baseline and report the denominator.
5. Record failure cases and scope limits beside the successful result.

From a paper claim to an engineering contract

The primary anchor for this chapter is Towards Highly Efficient Million-Token Context Intelligence from DeepSeek-V4. Reading a number from that source is only the first step. A reproducible contract has four layers:

Layer	Question to write down	Evidence
Mechanism	What operation, loss, state, or routing decision changes?	Equation, pseudocode, tensor shapes
Implementation	How is it realized on the named hardware and software stack?	Kernel, precision, layout, process groups
Measurement	Which denominator and baseline make the comparison fair?	Raw metrics, config, repeated runs
Scope	Where should the claim stop being trusted?	Failure cases, ablations, dated limitations

This separation prevents a frequent error in frontier-model writing: converting a theoretical reduction into a latency promise, or converting one internal benchmark into a universal quality ranking. The implementation can fail to realize the algorithm, and the workload can fail to expose the intended benefit.

Failure modes and misleading shortcuts

• V4 is no longer a hypothetical future release.
• The official arXiv submission is April 26, 2026.
• The attention mechanisms are CSA and HCA, not a generic full-plus-sparse description.
• Published FLOP ratios are not measured wall-clock speedups.
• 1M maximum context does not guarantee perfect retrieval at 1M.

These are not footnotes. Frontier-model engineering is dominated by boundary conditions: a method can be mathematically correct and still lose to memory traffic, data skew, numerical drift, evaluation leakage, or a poorly stated comparison. A credible result makes those boundaries visible.

How to audit claims about this topic

Rewrite each claim with its missing boundary: name the exact mechanism, identify the tensor or resource it changes, and attach the workload and measurement. Then construct a counterexample at the edge of the claim. If a sentence cannot survive that rewrite, treat it as orientation—not evidence.

Next, trace provenance. Prefer the primary report for configuration and results, the released code for implementation behavior, and your own profiler for product performance. Secondary explainers are valuable for intuition but should not silently become the source of a numerical claim.

Decision guide: when should you use this idea?

Use it when the bottleneck named in the thesis appears in profiler traces or controlled quality experiments, the necessary kernels and runtime support exist, and the added system complexity can be observed in production. Start with the smallest configuration that exposes the bottleneck.

Delay it when a dense or higher-precision baseline does not yet converge, the evaluation harness is unstable, or the claimed resource is not limiting the workload. Sophisticated architecture cannot compensate for an invalid baseline.

Reject it when its benefit exists only under a denominator irrelevant to the product—for example, theoretical FLOPs while user latency worsens—or when numerical, safety, or operational regressions exceed the measured gain.

Hands-on study lab

1. 1. Read the official config.
2. 2. Draw the hybrid attention paths.
3. 3. Separate model and serving claims.
4. 4. Build a V3.2-to-V4 fact table.

For each exercise, save the configuration, a tiny deterministic fixture, the raw measurements, and one failed case. The goal is not merely to make the code run; it is to make the conclusion independently checkable.

Quick self-check

What is the central idea?

V4 scales model capacity and context together by replacing uniform attention with specialized compressed paths and strengthening information flow through very deep sparse networks.

What is the most common reading mistake?

V4 is no longer a hypothetical future release.

What evidence should I demand?

An exact configuration, a fair baseline, primary-source support, end-to-end measurements, and failure cases at the limits of the claim.

How do I explain it to a new engineer?

Begin with the bottleneck, show one tiny worked example, trace the changed state, and only then introduce the official name. Finish by naming one situation where the method will not help.

How do I review an implementation?

Check indexing and masks, parameter sharing, dtype transitions, layouts, process-group scope, raw metric denominators, and behavior under an adversarial or worst-case fixture. A passing happy-path shape test is not enough.

Teach-back synthesis

Close the page and reconstruct the argument in five sentences: the bottleneck; the mechanism; the state or tensor that changes; the fair measurement; and the main failure mode. Then reopen the page and compare. If you can repeat the feature names but cannot state those five sentences, revisit the worked example.

Finally, connect the idea to two neighboring phases. DeepSeek's advantage is not one isolated invention: compressed attention changes the cache, sparse experts change active compute, FP8 changes arithmetic and bandwidth, distributed schedules hide communication, and reasoning training spends the resulting capacity differently. The series becomes useful when those dependencies form one mental model.

Key takeaways

• The official 2026 preview: V4-Pro and Flash, CSA+HCA hybrid attention, mHC, Muon, 32T+ tokens, reasoning modes, and million-token efficiency.
• The mechanism, training recipe, runtime implementation, and measured product behavior are separate layers of evidence.
• Numbers remain meaningful only with their workload, precision, hardware, context length, and date attached.
• A small reproducible test is more valuable than a large uncheckable diagram.

Primary sources and further reading

• DeepSeek-V4. Towards Highly Efficient Million-Token Context Intelligence.
• DeepSeek-V3.2. Pushing the Frontier of Open Large Language Models.
• DeepSeek-V3. DeepSeek-V3 Technical Report.

Source note: explanations and worked examples here are original. Numerical claims are scoped to the linked reports; rapidly changing model comparisons are dated in the article itself.