Humanity's Last ExamVerified
39.2%
HLE (AA run)
Models / DeepSeek
DeepSeek V4.1 Flash
DeepSeek V4.1 · Released 2026-09-10
DeepSeek's 10 Sep 2026 Flash refresh: 552B MoE, 8B active on prefill and 16B on decode, native vision, 1M context, 384K max output. API id deepseek-flash. Official off-peak list is $0.15 / $0.60 per 1M tokens (cache-miss input / output); peak doubles. Retired V4-Flash aliases route here.
Current DeepSeek Flash API model for cheap multimodal and agent work. Prefer official pricing and architecture notes over vendor eval tables; AA rows land when the snapshot lists the slug.
Composite56
Context1M
Input / 1M$0.15
Output / 1M$0.60
Knowledge cutoffunpublished
Output speed220 t/s
Statussolid
Family profile
Best published score in each covered benchmark family.
58/100 avg
- Reasoning
- 39
- Coding
- 52
- Long context
- 84
3 tested benchmark families
Editor's note
Recorded from DeepSeek's 10 Sep 2026 launch post, the live API pricing table (18 Sep 2026), and Hugging Face model card (552B MoE, CED 8B/16B active, 1M context, MIT weights). The Sep 10 note that deepseek-v4-pro would route here on 14 Sep was withdrawn; Pro remains a separate API. Vendor Terminal-Bench / DeepSWE / AutomationBench figures stay off the ledger.Benchmark placements
Per-benchmark positions
Where DeepSeek V4.1 Flash places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.
| Benchmark | Family | Rank | Score | Source | Date |
|---|---|---|---|---|---|
| AA-LCR | Long context | #6/ 184 | 84 | Artificial Analysis | 2026-09-28 |
| AA output speed | Performance | #9/ 191 | Artificial Analysis | 2026-09-28 | |
| SciCode | Coding | #37/ 97 | 52 | Artificial Analysis | 2026-09-28 |
| Humanity's Last Exam | Reasoning | #47/ 186 | 39 | Artificial Analysis | 2026-09-28 |
Family context
Strongest rows in each family
The three highest-scoring models with pages in each capability family. Where DeepSeek V4.1 Flash shows up, it's highlighted.
ReasoningDeepSeek V4.1 Flash · 39
- 196
- 295
- 395
Math
- 152
- 252
- 350
CodingDeepSeek V4.1 Flash · 52
- 174
- 273
- 370
Long contextDeepSeek V4.1 Flash · 84
- 189
- 286
- 385
Newest receipts
4 recentLast 120 days
Benchmark rows added to the public ledger for DeepSeek V4.1 Flash in the last 120 days. Older rows live in the full table below.
| Benchmark | Family | Score | Source | Days ago |
|---|---|---|---|---|
| AA-LCR | Long context | 84% | Artificial Analysis | 13 |
| AA output speed | Performance | 220 t/s | Artificial Analysis | 13 |
| Humanity's Last Exam | Reasoning | 39.2% | Artificial Analysis | 13 |
| SciCode | Coding | 51.9% | Artificial Analysis | 13 |
Family coverage
Best score per family
A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.
Price vs. performance
Where DeepSeek V4.1 Flash sits on the scatter
Hover any dot for name, score, and input price. Keyboard: tab through the top twelve, or use the ranking below.
AnthropicOpenAIGoogleAlibabaMetaZhipuxAIDeepSeekMiniMaxXiaomiMoonshotOtherMeituanMistralNVIDIACohere
Full benchmark ledger
Every catalog benchmark for DeepSeek V4.1 Flash. Scores link to the original source; gaps mean no public row exists yet.
4 of 13 catalog benchmarks have a sourced row for DeepSeek V4.1 Flash.
Reasoning
Coding
SciCodeVerified
51.9%
SciCode (AA run)
Long context
AA-LCRVerified
84%
AA-LCR (AA run)
Performance
AA output speedVerified
220 t/s
Median output tokens/s (1k prompt, default provider)
Changelog
Version history
- Re-checked DeepSeek API docs: V4-Flash aliases still route here. The planned 14 Sep deepseek-v4-pro reroute was withdrawn; Pro keeps its own API and billing.
- DeepSeek-V4.1-Flash GA. API id deepseek-flash. Off-peak list $0.15 / $0.60 per 1M tokens. V4-Flash aliases retire onto this model.
Sources
Where the numbers come from
- DeepSeek V4.1 Flash announcement· checked 2026-09-18
- DeepSeek API models and pricing· checked 2026-09-18
- DeepSeek API changelog 10 Sep 2026· checked 2026-09-18
- Hugging Face DeepSeek-V4.1-Flash· checked 2026-09-12
Questions this model page answers
Every answer below is assembled from the dated fields on this page. Nothing is written separately for search.
What is DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash is a DeepSeek model: DeepSeek's 10 Sep 2026 Flash refresh: 552B MoE, 8B active on prefill and 16B on decode, native vision, 1M context, 384K max output. API id deepseek-flash. Official off-peak list is $0.15 / $0.60 per 1M tokens (cache-miss input / output); peak doubles. Retired V4-Flash aliases route here.
What is DeepSeek V4.1 Flash for?
Current DeepSeek Flash API model for cheap multimodal and agent work. Prefer official pricing and architecture notes over vendor eval tables; AA rows land when the snapshot lists the slug.
What does DeepSeek V4.1 Flash cost?
DeepSeek V4.1 Flash lists $0.15 input and $0.60 output per 1M tokens. Pricing last checked 2026-09-18.
How large is the DeepSeek V4.1 Flash context window?
DeepSeek V4.1 Flash publishes a 1M token context window.
What is the VerdictPal status of DeepSeek V4.1 Flash?
DeepSeek V4.1 Flash carries Solid status on the model atlas.
The Pack · Editorial newsletter
New tools in your inbox. Free.
One short email when a tool ships or changes status. No tracking, no third-party analytics. Unsubscribe in one click.