Skip to main content
AI Choice Engine

Model comparison

DeepSeek V4 Pro vs Step 5 Preview

DeepSeek against StepFun, compared on context, price, and verified benchmark results.

Catalog record checked September 26, 2026Individual provider fields may changeEvidence confidence: High — see receipts below

DeepSeek

DeepSeek V4 Pro

Frontier · Open weights

vs

StepFun

Step 5 Preview

Frontier

AI model capability comparison
SpecificationDeepSeek V4 ProStep 5 Preview
ProviderDeepSeekStepFun
TierFrontierFrontier
Context window1M1M
Max output384KNot verifiedUnverified
Input / 1M tokens$1.32Winner: $1
Output / 1M tokens$3.96Winner: $2.70
WeightsOpenClosed
Parameters1.6T total / 49B active (MoE)600B total / 27B active MoE (StepFun)
Reasoning levelslow, high, maxNot verifiedUnverified
Modalitiestexttext, image
LicenseMITNot disclosedUnverified
API model iddeepseek-v4-prostep-5-preview
ReleasedApril 24, 2026September 18, 2026
Artificial Analysis Intelligence Index [max] (2026-09-26)36Not verifiedUnverified
Artificial Analysis Intelligence Index (2026-09-26)Not verifiedUnverified43.7
SWE-bench Verified (2026-04-24)80.6Not verifiedUnverified
GPQA Diamond (2026-04-24)90.1Not verifiedUnverified
Humanity's Last Exam (2026-04-24)48.2Not verifiedUnverified
MMLU-Pro (2026-04-24)87.5Not verifiedUnverified

Prices are USD per million tokens at standard rates, excluding batch and caching discounts. Bold indicates the better figure where one is objectively better. Values we could not confirm from the provider are shown as “Not verified” rather than estimated.

Benchmark receipts

Where each score comes from, and how far it can be compared across models.

  • 2026-09-26
    Artificial Analysis

    Composite index blending reasoning, knowledge, and coding evals into one 0–100 score.

    Comparable with caveatOnly same-version scores are comparable. v4.3.2 was rebaselined (the top score fell from 66 on v4.1.1 to ~58), so v4.1.1 figures must not be compared with v4.3.2 figures — check measuredAt.

  • 2026-04-24
    DeepSeek V4 Pro HF model card (vendor, Think Max, exact match)

    Real GitHub issue resolution: does the model's patch pass the hidden tests.

    Comparable with caveatPost-audit vendor claims and pre-audit scores sit on different task trust levels; scaffolding (agent harness, compute budget) also dominates results. Never aggregate across scaffolds.

Pricing tiers

DeepSeek V4 Pro: Official peak rates $1.32/$3.96 per MTok; off-peak $0.66/$1.98 (cache-miss). Cache hit $0.044 peak / $0.022 off-peak. Peak hours are weekdays 01:00–04:00 and 06:00–10:00 UTC. Open weights (MIT). 1.6T/49B active MoE.

Step 5 Preview: $1.00/$2.70 per MTok on the StepFun platform (cache-hit input $0.05); StepFun lists it under Multimodal Reasoning Models and says output tokens include reasoning. No max-output figure is published. StepFun says open weights follow on 2026-10-15. Release date per Artificial Analysis's listing (StepFun's launch page is undated).

FrontierOpen weightsRecord checked September 26, 2026

DeepSeek V4 Pro

DeepSeek V4 Pro is DeepSeek's large open-weight (MIT) MoE with a 1M-token context — DeepSeek says the cheaper V4.1-Flash now beats it.

Best for

  • Cost-sensitive hosted agents
  • Open-weight deployments
  • High-volume coding

Watch out

On 2026-09-10 DeepSeek said it is phasing out V4-Pro (it later reversed the 2026-09-14 re-route and still serves it); plan new work on V4.1-Flash. Self-hosting needs datacentre VRAM.

FrontierRecord checked September 26, 2026

Step 5 Preview

Step 5 Preview is StepFun's new frontier reasoning model — a 1M-context, vision-input API preview that lands near Kimi K3 and GLM 5.3 on the Artificial Analysis index at a lower price.

Best for

  • Long-context work on a budget
  • Chinese + English
  • Early testing before the open-weight drop

Watch out

Preview: weights are promised for 2026-10-15 but not yet released, and its max output is unpublished and the launch benchmarks are vendor-run.

When the cheaper one wins

Step 5 Preview is cheaper on output at $2.70 per million tokens against $3.96 for DeepSeek V4 Pro — about 1.5×. Use the cheaper tier for classification, extraction, summarisation, and any task where the expensive model’s extra score does not change the accepted output. The expensive one only pays if your hardest task actually fails on the cheap tier. These are standard-tier API rates, excluding batch and cache discounts.

Run the model picker

Evidence confidence: High

How strong and complete the evidence behind this comparison is — not a prediction of which model is better.

  • Pricing verified on both sides — Input and output rates are verified for both models.
  • 3/5 core specs verified on both sides — Not published for at least one side: max output, reasoning levels.
  • 1 shared named benchmark with differing scores — Measured on: Artificial Analysis Intelligence Index.
  • Verified within the last 90 days — Newest catalog check was 2 days ago.
  • Both models carry source citations — Each side has at least two catalog sources on record.

Source receipts

Each catalog figure was checked against the provider or an independent second source on the date shown.

Related comparisons

Diving deeper on one model? DeepSeek V4 Pro · Step 5 Preview

Common questions

DeepSeek V4 Pro vs Step 5 Preview

Answered from the verified figures on this page rather than general guidance.

Is DeepSeek V4 Pro or Step 5 Preview cheaper for input?
Step 5 Preview is cheaper at $1 per million input tokens, against $1.32 for DeepSeek V4 Pro — roughly 1.3× the price. Output tokens usually dominate a real bill, so weigh the output rate more heavily than the input rate. These are base rates; DeepSeek V4 Pro has tiered pricing: Official peak rates $1.32/$3.96 per MTok; off-peak $0.66/$1.98 (cache-miss). Cache hit $0.044 peak / $0.022 off-peak. Peak hours are weekdays 01:00–04:00 and 06:00–10:00 UTC. Open weights (MIT). 1.6T/49B active MoE. Step 5 Preview has tiered pricing: $1.00/$2.70 per MTok on the StepFun platform (cache-hit input $0.05); StepFun lists it under Multimodal Reasoning Models and says output tokens include reasoning. No max-output figure is published. StepFun says open weights follow on 2026-10-15. Release date per Artificial Analysis's listing (StepFun's launch page is undated).
Is DeepSeek V4 Pro or Step 5 Preview cheaper for output?
Step 5 Preview is cheaper at $2.70 per million output tokens, against $3.96 for DeepSeek V4 Pro — roughly 1.5× the price. Output tokens usually dominate a real bill, so weigh the output rate more heavily than the input rate. These are base rates; DeepSeek V4 Pro has tiered pricing: Official peak rates $1.32/$3.96 per MTok; off-peak $0.66/$1.98 (cache-miss). Cache hit $0.044 peak / $0.022 off-peak. Peak hours are weekdays 01:00–04:00 and 06:00–10:00 UTC. Open weights (MIT). 1.6T/49B active MoE. Step 5 Preview has tiered pricing: $1.00/$2.70 per MTok on the StepFun platform (cache-hit input $0.05); StepFun lists it under Multimodal Reasoning Models and says output tokens include reasoning. No max-output figure is published. StepFun says open weights follow on 2026-10-15. Release date per Artificial Analysis's listing (StepFun's launch page is undated).
Which has the larger context window, DeepSeek V4 Pro or Step 5 Preview?
Both accept about 1M tokens of context, so document length will not decide between them.
Should I use DeepSeek V4 Pro or Step 5 Preview?
Both sit in the frontier tier, so the choice usually comes down to price and context rather than capability. DeepSeek V4 Pro suits cost-sensitive hosted agents; Step 5 Preview suits long-context work on a budget.
Can I self-host DeepSeek V4 Pro or Step 5 Preview?
DeepSeek V4 Pro publishes open weights, but self-hosting and third-party provider access depend on the licence, hardware, serving support, and availability. Step 5 Preview is a closed model whose supported access paths are controlled by its provider.

Next step

Choosing between them

The verified figures that separate this pair, computed from the catalog rather than restated boilerplate.

  • Input price: Step 5 Preview costs $1 per 1M tokens versus $1.32 for DeepSeek V4 Pro — a 1.3x difference at the headline tier.
  • Deployment: DeepSeek V4 Pro publishes weights you can self-host; the other is API-only.
Run the model pickerEstimate stack costAll model comparisons

Choosing a harness rather than a model — Cursor, Copilot, Claude Code, Windsurf, or Lovable? Compare agentic harnesses · Latest releases.