DeepSeek V4.1 Flash vs V4 Pro
Who's stronger, pricier, and how to choose
After DeepSeek shipped V4.1 Flash on 2026-09-10, its release note said "Tests by multiple parties put V4.1-Flash ahead of V4-Pro on performance, cost, speed & total runtime". V4 Pro was not retired: the changelog and pricing page confirm it continues with unchanged billing. This page separates the two tiers.
Updated 2026-09-19
Four key differences
Official multi-party tests
Release note: V4.1-Flash beats V4-Pro on performance/cost/speed/total runtime.
V4 Pro status
Changelog continues V4 Pro with unchanged billing (V4-Pro-0813).
Context/output
V4.1 Flash pricing page lists 1M context, 384K max output.
Price structure
Both are peak/off-peak with off-peak at half; Flash unit price is far lower.
What each is
V4.1 Flash is the smallest model of the new family—fast, cheap, native vision. V4 Pro (V4-Pro-0813) is the heavy-reasoning line. The official claim is Flash now beats Pro on several axes, yet Pro continues. Pick by latency/cost vs strongest reasoning.
The official 'Flash beats Pro' claim
The 2026-09-10 release note says 'Tests by multiple parties put V4.1-Flash ahead of V4-Pro on performance, cost, speed & total runtime. We're phasing out V4-Pro'. The changelog and pricing page then confirmed V4 Pro continues with unchanged billing. Defer to changelog + pricing: both remain; Flash is the new default DeepSeek pushes.
Timeline
On 2026-08-13 V4 Pro went GA (V4-Pro-0813), DeepSeek's heavy line.
On 2026-09-10 V4.1 Flash released; official claim it beats Pro; release note mentions phasing out V4-Pro.
The changelog's 2026-09-10 entry and the pricing page confirmed V4 Pro continues with unchanged billing; not re-routed.
Confirmed vs caution
✅ Confirmed
V4.1 Flash released 2026-09-10; release note says Flash ahead on several axes; changelog continues V4 Pro with unchanged billing; current versions DeepSeek-V4.1-Flash and DeepSeek-V4-Pro-0813.
⚠️ Caution
The release note once said v4-pro routes to Flash from 2026-09-14, but the changelog/pricing page confirmed v4-pro continues with unchanged billing. Defer to changelog + pricing. No third-party benchmarks.
Same vs differs
Same
Both official DeepSeek API, same base URL, thinking/non-thinking and tool calls, 1M context.
Differs
Flash is faster/cheaper and officially ahead on several axes; Pro is the heavy-reasoning line at a higher unit price. Same peak/off-peak structure, an order of magnitude apart.
How to choose
Low latency, high volume, budget-sensitive: V4.1 Flash (lower peak/off-peak, cheaper cache hits). Strongest reasoning/complex tasks at a higher unit price: V4 Pro. Both share one key and endpoint on QCode—switch by model id.
Comparing on QCode
deepseek-v4.1-flash and deepseek-v4-pro are both on sale with real usage, one key and one quota. Run the same requests on both to build your own cost/quality baseline. The new official id deepseek-flash is not currently listed on /models; whether it can be set follows /models.
FAQ
Is V4.1 Flash really better than V4 Pro?
The release note cites multi-party tests putting Flash ahead on performance/cost/speed/total runtime. That's the official claim; still validate on your own tasks.
Will V4 Pro be retired?
The release note once said 'phasing out', but the changelog and pricing page confirm V4 Pro continues after 2026-09-14 with unchanged billing.
How big is the price gap?
Both peak/off-peak with lower cache-hit price; Flash's unit price is far below Pro (use the pricing page's daily table).
Same context?
Pricing page lists V4.1 Flash 1M context / 384K max output; V4 Pro also 1M.
How to compare on QCode?
Set deepseek-v4.1-flash and deepseek-v4-pro, run the same requests on one key, compare latency/quality/cost.
Should I migrate to Flash?
If 'Flash ahead' holds for your use case and you want lower cost/latency, yes; if you rely on Pro's strongest reasoning, Pro is still on sale.
Sources
DeepSeek release note news260910, changelog and pricing page, captured 2026-09-18. Where the release note and changelog conflict on v4-pro, defer to changelog + pricing. QCode ids/usage from /models and 30-day stats.
One key to compare both
deepseek-v4.1-flash and deepseek-v4-pro share one QCode key and quota—switch by model id, priced at official rate × service fee.
Related
DeepSeek V4.1 Flash released guide
Full spec, peak/off-peak pricing and QCode setup.
DeepSeek V4 Pro 0813 guide
Current v4-pro positioning, pricing, setup.
DeepSeek model id changes
New id, temporary routing and the v4-pro routing saga since 09-10.
Separates confirmed from caution; defer to official docs; no leaked third-party benchmarks.