AI Trend Notifier
EN
← wiki

$ cat wiki/models/deepseek-v4-1-flash.md

DeepSeek V4.1-Flash

Compared with

Spec

AttributeValue
DeveloperDeepSeek
Released2026-09-10
Announced2026-09-10
Context window1,000,000
Pricing$0.15/M input (cache miss) · $0.003/M input (cache hit) · $0.60/M output, off-peak; peak rates double — see Pricing
LicenseMIT (open-weight)
AvailabilityDeepSeek API, Hugging Face (deepseek-ai/DeepSeek-V4.1-Flash)

Release Date

2026-09-10 (source).

Two prefetch candidates on the 2026-09-10 run named this model a day before it shipped — DeepSeek Flash 4.1 is already being tested via API and rolling out and Deepseek Has Soft Retired Deepseek V4 Pro — and that run recorded both and adopted neither, on the ground that no changelog entry, API model string, model card or technical report stood behind them. Both are now confirmed by the release (source). The rule cost this wiki one day and prevented it publishing a rumour; it is recorded here because a correct rumour is the case that makes the rule feel expensive.

Architecture — a Causal Encoder-Decoder, and the asymmetric activation is the point

DeepSeek describes the model as the first in a Causal Encoder-Decoder (CED) family (2 passes). Reported structure (source):

PropertyValue
Backbone parameters552B
Active during prefill8B
Active during decode16B
Layers40 — a 20-layer encoder and a 20-layer decoder (1 pass)
Cache constructionthe decoder builds its cache once from the encoder output, rather than one cache per layer (1 pass)
Input modalitiesnative image and text (1 pass)
KV cache890 bytes per token, stated as ~a quarter of DeepSeek-V4-Flash's (1 pass)
Checkpoint~475 GiB across 48 safetensors shards (1 pass)
**No other model on this wiki activates a different parameter count for reading than
for writing.** Every MoE page here — DeepSeek V4-Flash at 284B/13B,
Hy4 preview at 770B/49B, GLM-5.3-Flash at 320B-A18B — quotes one
activation figure that covers both phases. The 8B/16B split is what the memory claim
rests on: fewer parameters are engaged while the prompt is being consumed, which is
where a long-context agentic workload spends most of its tokens.

What is not established is the total parameter count as distinct from the "552B backbone" figure, and whether the 1M window is the same for both halves (source).

Pricing

DeepSeek's published API rates. The off-peak column is the one three search passes agree on (source):

Per 1M tokensCache-hit inputCache-miss inputOutput
Off-peak$0.003$0.15$0.60
Peak$0.006$0.30$1.20
Peak hours are 01:00–04:00 and 06:00–10:00 UTC on weekdays; peak rates are stated
as double off-peak. The release is reported to cut Flash API prices by **roughly
11% to 57% depending on token type** (1 pass).

Cache-hit input at $0.003/M off-peak stands against $0.022/M for the outgoing V4-Pro (1 pass) — a cell that matters more than the headline rate for the agentic workloads this model is aimed at, since a long-running agent re-reads its own context on every turn.

Note against DeepSeek V4-Flash. That page's schedule is quoted in RMB, because every source read for the 2026-08-16 change gave RMB only, and it explicitly declines to convert. This page's figures are reported in USD by every source read. The two schedules are therefore not directly comparable on this wiki, and no conversion is performed here either.

V4-Pro is being retired into this model

From 04:00 UTC on 2026-09-14, every request to deepseek-v4-pro is routed to V4.1-Flash and billed at Flash rates, reported as roughly a quarter of V4-Pro's, and the routing continues until V4.1-Pro launches (2 passes) (source).

V4.1-Pro is named as forthcoming and carries no date, size, price or benchmark in anything read. It is a watch item, not a page.

The affected page is DeepSeek V4-Pro-0813, whose Availability row is updated to record the routing.

Benchmarks

Vendor-stated, no evaluation harness named for any row, and no independent measurement of any kind appears in anything read — no Artificial Analysis index, no leaderboard placement (source).

BenchmarkV4.1-FlashClaude Opus 5GPT-5.6 Sol
Terminal-Bench 2.190.6unknownunknown
DeepSWE v1.174.274.073.0
AutomationBench54.850.345.8
Humanity's Last Exam36.856.3unknown
ProgramBench20.337.0unknown
**The margins on the top three rows are 0.2, 1.2 and 4.5 points, and the margins on
the bottom two are 19.5 and 16.7 against it.** This is a model that leads on agentic
and terminal work and trails badly on breadth — and the release's own comparison set
is the one that shows both. Per Eval Harness Configuration, vendor-run
agentic figures with no published harness are claims about a (model, harness) pair,
not model properties; a 0.2-point lead over Opus 5 on DeepSWE is inside the range a
scaffold choice moves.

One extract states that at max reasoning effort V4.1-Flash "consistently outperformed the V4-Pro" (1 pass); no table for that comparison was read.

Use Cases

  • Long-running agentic and terminal work, where the cache-hit rate and the 890 bytes/token cache footprint compound over a session
  • Cost-sensitive serving replacing V4-Pro, which is being routed here anyway
  • Self-hosting under MIT — though at ~475 GiB the checkpoint is not a single-consumer-GPU proposition, unlike Muse Glimmer

Compared To

ModelOpen?Active paramsTerminal-Bench 2.1
DeepSeek V4.1-FlashYes (MIT)8B prefill / 16B decode90.6 (vendor)
DeepSeek V4-Pro-0813Yes (MIT)49B87.9 (vendor)
DeepSeek V4-FlashYes (MIT)13B82.7 (vendor)
Hy4 previewYes (Apache 2.0)49B85.4 (vendor)
GLM-5.3-FlashYes (MIT)18B84.3 (vendor)
Every figure in that column is vendor-stated on an unnamed harness, so the column
orders claims, not models.

Conflicting Reports

  • Off-peak pricing. One search pass gave off-peak figures of $0.075/M for a cache-miss input and $0.30/M for output — exactly half the three-pass figures in both cells, which is the shape of an extract that applied the off-peak discount twice. The three-pass figures are carried above; the single-pass pair is recorded here and not adopted (source).

Referenced by

Sources