# DeepSeek V4.1 Flash: Release Date, the Test Model That Expires September 10, and What DeepSeek Has Published

> DeepSeek V4.1 Flash is unreleased as of Sept 8, 2026. A test id that expires Sept 10 is live for API users. Every claim dated, plus DeepSeek's real release cadence.

- Author: Nitish Garg, Founder & CEO, CellCog
- Published: 2026-09-08
- Canonical (HTML): https://cellcog.ai/blog/deepseek-v4-1-flash-release-date/
- Section: Guides / Choosing a platform
- Publisher: CellCog (https://cellcog.ai), the AI employee platform. Blog index for agents: https://cellcog.ai/blog/llms.txt

## Key points

- DeepSeek V4.1 Flash is not released as of September 8, 2026. DeepSeek's API news page, its model list, its change log, its Hugging Face organization and its X account contain no announcement, model card, price or spec for a 4.1 model.
- What exists is a test. On September 8 at about 3 pm Beijing time, DeepSeek posted a notice in its official community group opening an 'intermediate version' of V4.1 Flash under the API model id deepseek-v4.1-flash-expires-on-0910, billed at V4-Flash rates, limited to 20 concurrent requests per account. The id says it expires September 10.
- The notice, as relayed by IT之家, Phoenix Tech and Machine Heart, describes a new model structure, native multimodal support, more capability, more speed and lower cost. None of that is in a DeepSeek document, and no parameter count, context window or benchmark has been published.
- The most telling detail is a question in DeepSeek's feedback form: whether the V4.1 Flash intermediate version can fully replace the online DeepSeek V4 Pro. A Flash-class model replacing Pro would be the story if it happens.
- Speed claims of about 400 tokens per second and 3.9 to 6 times faster than V4-Flash-Vision-Exp come from one X user's self-tests, relayed by press. They are unverified.
- DeepSeek's own cadence says a release is plausible soon: V4 Preview April 24, V4-Flash-0731 on July 31, V4-Pro-0813 on August 13, V4-Flash-Vision-Exp on August 21. The last three landed 13 and 8 days apart.
- This is a living tracker. When DeepSeek publishes an announcement, a model card or a price, the confirmed facts land here the same day next to a scorecard of every claim above.

## At a glance

- **Is DeepSeek V4.1 Flash released?** No. As of September 8, 2026 there is no DeepSeek announcement, model card, permanent API id, price or benchmark for V4.1 Flash. What is live is a test id, deepseek-v4.1-flash-expires-on-0910, opened to API users through a community-group notice and named to expire on September 10.
- **When is DeepSeek V4.1 Flash coming out?** DeepSeek has not said. The test id's expiry on September 10 is the only date attached to the model, and it is an end date for a test, not a launch date. DeepSeek's recent releases have arrived 8 to 13 days apart, so a release in September would fit its pattern.
- **What is different about V4.1 Flash?** According to the relayed notice: a new model structure, native multimodal support, stronger capability, faster speed and lower cost. DeepSeek's April V4 announcement described the V4 structure as token-wise compression plus DeepSeek Sparse Attention with a 1M context; what changed in 4.1 has not been published.
- **What is DeepSeek's current lineup?** Three API models as of September 8: deepseek-v4-flash (serving DeepSeek-V4-Flash-0731, released July 31), deepseek-v4-pro (serving DeepSeek-V4-Pro-0813, released August 13) and the experimental deepseek-v4-flash-vision-exp (released August 21). V4-Flash is 284B total and 13B active parameters with a 1M context, per DeepSeek's April announcement.

DeepSeek V4.1 Flash has a date attached to it, and the date is an expiry. On September 8, 2026, at about 3 pm Beijing time, DeepSeek posted a notice in its official user community group: an "intermediate version" of V4.1 Flash was open for testing under the API model id `deepseek-v4.1-flash-expires-on-0910`, billed at the same rate as `deepseek-v4-flash`, capped at 20 concurrent requests per account. The notice, as relayed by Chinese technology press within the hour, promised a new model structure, native multimodal support, more capability, more speed and lower cost. It did not promise a release date, and DeepSeek has published nothing since.

We checked the channels DeepSeek actually uses. As of September 8, 2026, the [API news page](https://api-docs.deepseek.com/news/) ends at the August 21 release of V4-Flash-Vision-Exp. The [models and pricing page](https://api-docs.deepseek.com/quick_start/pricing) lists three models, none of them 4.1. The [change log](https://api-docs.deepseek.com/updates) has no 4.1 entry. The [deepseek-ai organization on Hugging Face](https://huggingface.co/deepseek-ai) has no 4.1 repository, and the newest model there is still [V4-Flash-Vision-Exp](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp). The company's X account has said nothing. DeepSeek's own April announcement carries a line worth keeping in mind here: "please rely only on our official accounts for DeepSeek news."

So this page does what our [Qwen 4](https://cellcog.ai/blog/qwen-4-release-date/), [Gemini 4](https://cellcog.ai/blog/gemini-4-release-date/) and [Grok 4.7](https://cellcog.ai/blog/grok-4-7-release-date/) trackers do: every claim dated, the company's documents separated from the relayed notice and both separated from community tests, updated the same day the story moves. DeepSeek is a different kind of tracker than Grok. There is no founder posting dates, no leaked benchmark sheet, no Reddit war. There is a model id with a two-day life and a release history that is unusually regular. That history is most of what we can say.

## What we actually know

*Table: DeepSeek V4.1 Flash claims vs verifiable status, September 8, 2026*

| Claim | Status |
|---|---|
| A V4.1 Flash test model is live on the API | Confirmed: id `deepseek-v4.1-flash-expires-on-0910`, per DeepSeek's community-group notice as relayed by IT之家, Phoenix Tech and Machine Heart |
| Billing equals `deepseek-v4-flash` | Per the notice, relayed; DeepSeek's pricing page does not list the id |
| 20 concurrent requests per account | Per the notice, relayed |
| The test ends September 10 | The model id says so; DeepSeek has not published a date |
| New model structure | Per the notice; nothing published on what changed |
| Native multimodal support | Per the notice; V4-Flash-Vision-Exp already accepts images via an added encoder |
| Stronger, faster, cheaper | Per the notice; no benchmark, throughput or price published |
| About 400 tokens per second | Community self-test relayed by Machine Heart; unverified |
| 3.9 to 6 times faster than V4-Flash-Vision-Exp | One X user's self-tests (@NFT_Chen) relayed by Machine Heart; unverified |
| "Can it replace the online V4 Pro?" | A question in DeepSeek's own tester feedback form, per IT之家 |
| Release date | Nothing. DeepSeek has not announced one |
| Parameters, context window, permanent API id, model card, price | Nothing exists |
| DeepSeek's current lineup | `deepseek-v4-flash` (V4-Flash-0731), `deepseek-v4-pro` (V4-Pro-0813), `deepseek-v4-flash-vision-exp`, per the pricing page |

## What the notice said, and who carried it

DeepSeek did not publish the notice on a public page, so the primary record is the press that quoted it. [IT之家](https://finance.sina.com.cn/tech/digi/2026-09-08/doc-inirceww9427669.shtml) posted at 16:05 Beijing time, citing readers who saw the message in DeepSeek's official group: an intermediate version of DeepSeek V4.1 Flash is in internal testing, it uses a new model structure with native multimodal support, it is stronger, faster and cheaper, testers keep `base_url` unchanged and set the model to `deepseek-v4.1-flash-expires-on-0910`, billing matches `deepseek-v4-flash`, and each account is limited to 20 concurrent requests. IT之家 also reported the feedback form's question about replacing V4 Pro. [Phoenix Tech](https://i.ifeng.com/c/8wFovW2dxl1) carried the same points at 16:17. [Machine Heart](https://finance.sina.com.cn/tech/roll/2026-09-08/doc-inirceww4639322.shtml), at 18:00, put the group message at about 3 pm and added that "internal test" was generous: any API user could add the model name and call it.

Three things about the notice are worth reading carefully. First, "intermediate version" is DeepSeek's own framing. This is not the release model; it is a checkpoint offered for feedback, and the expiry in the id enforces that. Second, "new model structure" is a stronger claim than DeepSeek made for V4-Flash-0731, whose [model card](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731) describes it as the official release of V4-Flash with enhanced agentic capabilities and the same structure as the DSpark variant. A structural change between 4 and 4.1 is a bigger step than the version number suggests. Third, "native multimodal" is a distinction from the current approach. [V4-Flash-Vision-Exp](https://cellcog.ai/blog/deepseek-v4-flash-vision-exp/) added a vision encoder and aligner to the text model and kept training. Native means the images are part of the base design. If true, it explains why DeepSeek would call it a new structure.

## The rumor chain, dated

The chain is short, because DeepSeek does not leak the way Western labs do. What it does instead is ship with the date in the file name.

*Table: The DeepSeek V4-era chain, dated*

| Date | Event | Source type |
|---|---|---|
| Apr 24 | DeepSeek-V4 Preview: V4-Pro 1.6T/49B active, V4-Flash 284B/13B active, 1M context, open weights | DeepSeek, official |
| Jul 31 | DeepSeek-V4-Flash-0731 official release, superseding the preview | DeepSeek, official |
| Aug 13 | DeepSeek-V4-Pro GA (V4-Pro-0813) with DeepSeek Harness; peak and off-peak API pricing announced | DeepSeek, official |
| Aug 16 | New pricing takes effect, 16:00 UTC | DeepSeek, official |
| Aug 21 | DeepSeek-V4-Flash-Vision-Exp live on the API; Harness 0.1.1 | DeepSeek, official |
| Aug 31 | V4-Flash-Vision-Exp open weights on Hugging Face, MIT license | DeepSeek, official |
| Sep 1 to 7 | DeepSeek Harness 0.1.3 pre-release builds tagged on GitHub | DeepSeek, official (pre-release) |
| Sep 8, 3 pm CST | Community-group notice: V4.1 Flash intermediate version, `expires-on-0910` id | DeepSeek notice, relayed by press |
| Sep 8, 16:05 CST | IT之家 reports the notice and the V4 Pro replacement question | Press |
| Sep 8, 18:00 CST | Machine Heart reports community speed tests, 400 tokens per second | Press, citing users |
| Sep 8 | DeepSeek news page, pricing page, change log, Hugging Face, X: no "4.1" | Our own read |
| Sep 10 | Test id expires, per its name | Implied by DeepSeek's id |

**The expiry is the only date, and it is not a launch date.** `expires-on-0910` tells testers when the checkpoint disappears. It does not tell anyone when a release model arrives, and DeepSeek's dated suffixes have always described the release itself (0731, 0813), never a countdown. Reading September 10 as a release date is the mistake this page exists to prevent.

**The speed numbers are one person's tests.** Machine Heart credits the 400 tokens per second figure to unnamed community members and the 3.9 to 6 times speedups to X user @NFT_Chen, comparing the test model against V4-Flash-Vision-Exp on a 49K-token retrieval task (5.2 times) and SVG code generation (6 times). Those are plausible for a model DeepSeek itself calls faster, and they are self-tests of an intermediate checkpoint over a two-day window. We list them as reported. The [Gemini 3.8 Flash](https://cellcog.ai/blog/gemini-3-8-flash/) comparison Machine Heart draws is the comparison DeepSeek's launch table will eventually make or avoid; until then it is a tweet.

**The V4 Pro question is the real signal.** DeepSeek asked its testers whether this Flash-class model can fully replace V4 Pro online. V4-Pro is the 1.6T total, 49B active flagship; V4-Flash is 284B total, 13B active. If DeepSeek is seriously weighing a Flash replacing Pro in its own service, the new structure is doing a great deal of work, and it would fit a company whose April announcement led with "cost-effective 1M context" rather than raw scale.

## DeepSeek's cadence

The reason a September release is plausible has nothing to do with the notice. It is the calendar.

*Table: DeepSeek releases since V3 and the gap since the previous release*

| Model | Announced by DeepSeek | Days since previous |
|---|---|---|
| DeepSeek-V3 | December 26, 2024 | n/a |
| DeepSeek-R1 | January 20, 2025 | 25 |
| DeepSeek-V3-0324 | March 25, 2025 | 64 |
| DeepSeek-R1-0528 | May 28, 2025 | 64 |
| DeepSeek-V3.1 | August 21, 2025 | 85 |
| DeepSeek-V3.1-Terminus | September 22, 2025 | 32 |
| DeepSeek-V3.2-Exp | September 29, 2025 | 7 |
| DeepSeek-V3.2 | December 1, 2025 | 63 |
| DeepSeek-V4 Preview | April 24, 2026 | 144 |
| DeepSeek-V4-Flash-0731 | July 31, 2026 | 98 |
| DeepSeek-V4-Pro-0813 | August 13, 2026 | 13 |
| DeepSeek-V4-Flash-Vision-Exp | August 21, 2026 | 8 |
| DeepSeek V4.1 Flash | Test id September 8; no release | 18 and counting |

Dates come from the [DeepSeek API news index](https://api-docs.deepseek.com/news/) and change log; V4-Flash-0731 is dated by its July 31 change log entry, and V3-0324 by its news page rather than the March 24 in its name. Two readings of the table. The long gaps are between generations: 144 days from V3.2 to V4 Preview, 98 from the preview to the first final V4 model. The short gaps are inside a generation: 13 days from V4-Flash-0731 to V4-Pro, 8 from V4-Pro to Vision-Exp. A 4.1 that is a structural change sits somewhere between those two rhythms. The September 8 test is 18 days after Vision-Exp; if V4.1 Flash ships in September, it will be the fastest structural update DeepSeek has made, and the expiring test id is the first time DeepSeek has previewed a model this way rather than shipping it.

## What 4.1 would land on top of

The V4 lineup is well documented, which is what makes the 4.1 silence stand out. DeepSeek's [April 24 announcement](https://api-docs.deepseek.com/news/news260424/) gave the architecture: V4-Pro at 1.6T total and 49B active parameters, V4-Flash at 284B total and 13B active, both with a 1M context as the default across all official DeepSeek services, built on token-wise compression plus DeepSeek Sparse Attention. The July 31 [V4-Flash-0731 card](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731) calls itself the official release of V4-Flash and reports it beating the V4-Pro preview on agent benchmarks despite the far smaller active count (Terminal Bench 2.1 at 82.7, DeepSWE at 54.4, Toolathlon-Verified at 70.3 on DeepSeek's own table). The card's weight total reads about 304B because the release ships with a speculative decoding module attached; the announced base is 284B. The [August 13 V4-Pro GA post](https://api-docs.deepseek.com/news/news260813/) added reasoning effort levels for both models, native OpenAI Responses API support, and the pricing change: peak and off-peak rates, off-peak 50 percent lower, effective August 16 at 16:00 UTC. Current figures are on the [pricing page](https://api-docs.deepseek.com/quick_start/pricing); the test id is billed at whatever `deepseek-v4-flash` costs at the hour you call it.

The agent tooling is moving too. DeepSeek Harness 0.1.1 shipped with Vision-Exp on August 21, and the [GitHub repository](https://github.com/deepseek-ai/deepseek-harness) tagged 0.1.3 pre-release builds every day from September 1 to 7. A new model with a new structure arriving alongside a harness release would match how V4-Pro landed. Our record of the wider open-weight wave this lands in is in the [Qwen3.8-Flash-Next](https://cellcog.ai/blog/qwen3-8-flash-next/) and [GLM-5.3-Flash](https://cellcog.ai/blog/glm-5-3-flash/) posts.

## What we are watching for

We run production work on three model families and read the open-weight labs closely, so this list is an operator's.

- **An entry on the API news page.** Every DeepSeek release since V3 has one. The moment a 4.1 page appears at api-docs.deepseek.com/news, this stops being a rumor page.
- **A permanent model id.** `deepseek-v4.1-flash` without an expiry, or a dated release suffix like the ones on 0731 and 0813, on the models and pricing page.
- **Weights and a card on Hugging Face.** V4-Flash-Vision-Exp's weights followed its API release by ten days. A card would give the first parameter count and confirm or retire the "new structure" line.
- **September 10.** If the test id expires and nothing replaces it, that is a scorecard row. If a second test id appears with a later expiry, that is a pattern.
- **A price.** "Lower cost" is in the notice. DeepSeek's V4 pricing change was announced eight days before it took effect; a 4.1 price would likely arrive the same way.
- **The V4 Pro decision.** Whether DeepSeek's release post says anything about the online service's default model, and whether `deepseek-v4-pro` keeps pointing at 0813.
- **Multimodal spec.** Whether "native" means the release model accepts images without a separate vision-exp id, which would retire the August 21 experiment.

## The tracker

This page updates when facts change, not when posts get louder. As of September 8, 2026: DeepSeek V4.1 Flash exists as a test model id with a two-day life, described in a group notice, tested by strangers and announced nowhere. No release date, no spec, no benchmark and no price exist in any DeepSeek document. The likeliest picture, on DeepSeek's own rhythm, is a release in September with the date in its name. When DeepSeek moves, the confirmed facts and a graded scorecard of every claim above land here the same day.

## Update log

This is a living page; when the story moves, the update lands here.

As of September 8, 2026: page opened. Test id `deepseek-v4.1-flash-expires-on-0910` live per DeepSeek's community notice; no official announcement to log.

## FAQ

**Has DeepSeek announced V4.1 Flash?**

Not on any channel it uses for announcements. As of September 8, 2026 the last entry on the DeepSeek API news page is the August 21 release of DeepSeek-V4-Flash-Vision-Exp, the pricing page lists three models with no 4.1, the change log has no 4.1 entry, the deepseek-ai organization on Hugging Face has no 4.1 repository, and the company's X account has posted nothing about it. The only DeepSeek-authored text about V4.1 Flash is a notice in its official user community group on September 8, which was relayed by Chinese technology press within an hour.

**What does deepseek-v4.1-flash-expires-on-0910 mean?**

It is a temporary API model id. DeepSeek's notice told testers to keep their base_url and set the model name to deepseek-v4.1-flash-expires-on-0910. Billing is the same as deepseek-v4-flash and each account is limited to 20 concurrent requests. The suffix is the expiry: the id stops working on September 10. DeepSeek has used dated suffixes before for release versions, such as V4-Flash-0731 and V4-Pro-0813, but an 'expires-on' suffix is new and it marks a test window, not a release.

**Is V4.1 Flash faster than V4-Flash?**

Reported, not confirmed. Machine Heart's September 8 write-up cites community users reaching about 400 tokens per second and one X user, @NFT_Chen, measuring 3.9 to 6 times faster end-to-end than V4-Flash-Vision-Exp, including 5.2 times faster on a 49K-token retrieval task and 6 times faster on SVG generation. Those are self-tests on a test model during a two-day window. DeepSeek has published no throughput figure. If the release model ships with a spec sheet, the speed row goes into the tracker with its source.

**Will V4.1 Flash replace V4 Pro?**

DeepSeek is asking that question itself. Its tester feedback form, per IT之家's report, asks whether the V4.1 Flash intermediate version can fully replace the online DeepSeek V4 Pro. That would be unusual: V4-Pro is DeepSeek's 1.6T total, 49B active flagship and V4-Flash is the 284B total, 13B active efficient model. A Flash-class model good enough to retire Pro from the online service would say more about the new structure than any benchmark. Nothing has been decided that DeepSeek has published.

**What is DeepSeek's release pattern?**

Fast and dated, with the dates in the model names. In 2026 alone: V4 Preview on April 24, V4-Flash-0731 on July 31, V4-Pro-0813 with the DeepSeek Harness on August 13, V4-Flash-Vision-Exp on the API on August 21 and its open weights on August 31. The three summer releases landed 13 and 8 days apart. On that rhythm a September V4.1 release is plausible; on DeepSeek's history, a test model with an expiry date has not always turned into a release of the same name.

**Should I wait for DeepSeek V4.1 Flash?**

If you use DeepSeek's API, the test id is available now at V4-Flash pricing and vanishes on September 10, so trying it costs nothing beyond tokens. If you are choosing where your agents run, the model underneath matters less than the layer around it: memory that survives the conversation, a task board, an inbox, approvals a human can see. CellCog does not route to DeepSeek today; our tiers run on Gemini 3.8 Flash, Claude Fable 5.1 and Opus 5, and a standing AI employee keeps its role across every engine swap. Plans start at $8 a month; you pay for the work, not the hire, and the cost depends purely on how much work you assign.

## Related

- [DeepSeek Opens V4-Flash-Vision-Exp: MIT Weights, Real Specs, and the Multimodal Agent Bet](https://cellcog.ai/blog/deepseek-v4-flash-vision-exp/index.md)
- [Qwen 4: Release Date, Leaks, and the Architecture Alibaba Already Shipped](https://cellcog.ai/blog/qwen-4-release-date/index.md)
- [GLM-5.3-Flash Is Ox Alpha: The Reveal, the Specs, and the Real Pricing](https://cellcog.ai/blog/glm-5-3-flash/index.md)
- [Gemini 4: Release Date, Leaks, and What Google Has Actually Said](https://cellcog.ai/blog/gemini-4-release-date/index.md)

---

Markdown alternate of https://cellcog.ai/blog/deepseek-v4-1-flash-release-date/. Try CellCog free, no credit card needed: https://cellcog.ai/signup