ModelsCompareBest forBenchmarksStatusPricingAPI
Compare / head-to-head

DeepSeek V4 ProvsStep 3.5 Flash

DeepSeek V4 Pro leads 1 of 2 shared benchmarks. DeepSeek V4 Pro has the larger context window (1M tokens).

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
1 – 0
DeepSeek V4 Pro leads · 1 tied
Cheaper per token
list price, input + output
Larger context
DeepSeek V4 Pro
1M tokens
Provider uptime (30d)
100% · —
DeepSeek · StepFun
DeepSeek V4 Pro
DeepSeek · released 2026-08-13
text
Context
1M
Max out
384K
Input /1M
$0.435
Output /1M
$0.87
Cached /1M
$0.004
Scores
11 · 9 core
Step 3.5 Flash
StepFun · released 2026-02-12
text
Context
256K
Max out
Input /1M
Output /1M
Cached /1M
Scores
18 · 5 core
Quality

Benchmark matrix

BenchmarkDeepSeek V4 ProStep 3.5 FlashΔEdge
Reported by both · 2
AIME 202696.7% matharena96.7% matharena0Tie
LMArena Elo1450.6 1403.7 +46.9DeepSeek V4 Pro
Only DeepSeek V4 Pro reports · 9
GPQA Diamond90.9% epochnot reported
DeepSWE v1.162.7% not reported
FrontierMath Tier 4 v2 (Epoch AI run)2.4% epochnot reported
FrontierMath Tiers 1-3 v2 (Epoch AI run)45.3% epochnot reported
Humanity's Last Exam (with tools)60% not reported
OTIS Mock AIME 2024-2025 (Epoch AI run)96.7% epochnot reported
SimpleQA Verified47% epochnot reported
SWE-bench Verified (Epoch AI run)77.6% epochnot reported
Terminal-Bench 2.187.9% not reported
Only Step 3.5 Flash reports · 15
SWE-bench Verifiednot reported74.4%
AIME 2025not reported97.3%
BrowseCompnot reported51.6%
BrowseComp (with context manager)not reported69%
BrowseComp-ZHnot reported66.9%
BrowseComp-ZH (with context manager)not reported73.7%
GAIA (no file)not reported84.5%
HMMT 2025 (Feb/Nov average)not reported96.2%
IMOAnswerBenchnot reported85.4%
LiveCodeBench v6not reported86.4%
ResearchRubricsnot reported65.3%
Tau2-Benchnot reported88.2%
Terminal-Bench 2.0not reported51%
xbench-DeepSearch 2025.05not reported83.7%
xbench-DeepSearch 2025.10not reported56.3%
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is DeepSeek V4 Pro minus Step 3.5 Flash in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecDeepSeek V4 ProStep 3.5 FlashEdge
Context window1M tokens256K tokensDeepSeek V4 Pro
Max output384K tokens
Input price / 1M$0.435
Output price / 1M$0.87
Cached input / 1M$0.004
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
$1.31
ModalitiestexttextTie
Released2026-08-132026-02-12
Cited benchmark scores1118
Reliability

Provider status

All providers
More matchups

DeepSeek V4 Pro vs …

More matchups

Step 3.5 Flash vs …

Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run DeepSeek V4 Pro and Step 3.5 Flash on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.