Skip to content
llm-spend
GitHub
Cost anatomy

DeepSeek-V4.1 Flash

DeepSeek · DeepSeek direct APIDirect API

Fresh input / cached input / output cost breakdown, active and scheduled rates, provenance, same-model deployment markup, and cost-comparable alternatives for this purchasable lane.

Start with a pattern

Illustrative workloads

Pick a useful starting shape, then fine-tune every input below.

Fine-tune

Workload and pricing scenario

Adjust the token shape, cache behavior, time window, and service tier. Every number below updates immediately.

Interactive

Workload cost calculator

60M
210K
90%

Share of input tokens served from cache. Applied only to models with a cache meter.

Time

Rows with a rate that changes by time of day, promo window or service tier are priced for this scenario instead of their flat listed rate; a brand-colored label under the model name names which one applies. Context size for band-priced rows uses the input-token count below.

Picking a specific hour also previews time-of-day rates that are scheduled but have not started billing yet; those rows are labelled · preview with their start instant. Now only ever shows rates that are billable at this moment.

Rate state

What this lane charges right now

Priced under the workload and scenario above. The full published schedule below always shows every variant regardless of which scenario is selected.

Off-peak — priced under the selected scenario, differs from the flat listed rate

Input / 1M
$0.15
CHF 0.121
Cached input / 1M
$0.003
CHF 0.002
Output / 1M
$0.60
CHF 0.483
Full published schedule
Off-peakChanges in 2d 8h
Peak$0.30 / $0.006 / $1.20Off-peak$0.15 / $0.003 / $0.60Peak (legacy V4 Flash)$0.44 / $0.014 / $1.32Off-peak (legacy V4 Flash)$0.22 / $0.007 / $0.66
Cost anatomy

Where this workload's cost goes

60M input tokens (90% cache hit) and 210K output tokens, at the resolved rate above.

$0.90
Fresh input · CHF 0.724
$0.162
Cached input · CHF 0.13
$0.126
Output · CHF 0.101
$1.19
Workload total · CHF 0.956

$0.90 + $0.162 + $0.126 = $1.19

Provenance

How much to trust these numbers

InputofficialCached inputofficialOutputofficial

The `deepseek-flash` endpoint now serves V4.1 Flash. Before 2026-09-10 04:00 UTC, this row's base and legacy variants preserve the V4 Flash schedule; from the transition onward, off-peak is $0.15/$0.003/$0.60 and weekday peak is $0.30/$0.006/$1.20 per M (cache hit / cache miss / output order is shown in the source table; the catalog fields are input / cached input / output).

DeepSeek's official API pricing page and September 10 V4.1 Flash announcement, re-read 2026-09-11: `deepseek-flash` is DeepSeek-V4.1-Flash with a 1M context and 384K maximum output; legacy `deepseek-v4-flash` names route to it. The base fields retain the pre-transition V4 Flash off-peak rate for deterministic historical resolution, while the variants below carry the current V4.1 Flash schedule.

Direct vs Foundry

Same-model deployment markup

How much more Microsoft Foundry charges for this exact model, over this Direct rate.

Not available — no Microsoft Foundry listing exists for this exact model in the catalog.

Cost-comparable

Other lanes within ±25% of this workload's cost

Same provider first, then ranked by cost proximity, for 60M in / 210K out. This is a cost ranking only — never a quality recommendation.

  • Qwen3.7 Flash
    Qwen · Model Studio (Intl)
    Direct API
    $1.22
    CHF 0.985
    +$0.036 (+3%) vs this lane