DeepSeek-V4.1-Flash: Specs, Pricing & API Model ID
DeepSeek-V4.1-Flash is DeepSeek’s Sept 10, 2026 API model — an asymmetric 552B-parameter MoE with a Causal Encoder–Decoder (about 8B active parameters on input, 16B on output), native multimodal/vision, 1M context and up to 384K max output. The live API model id is deepseek-flash. Official peak/off-peak prices (USD per 1M tokens): cache hit $0.003 / $0.006, cache miss $0.15 / $0.3, output $0.6 / $1.2. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday–Friday, excluding Chinese public holidays. Legacy ids deepseek-v4-flash and deepseek-v4-flash-vision-exp route to V4.1-Flash; from Sept 14, 2026 04:00 UTC, deepseek-v4-pro also routes to Flash at Flash rates until V4.1-Pro. This page summarises DeepSeek’s own docs only.
DeepSeek-V4.1-Flash at a glance
| DeepSeek-V4.1-Flash | |
|---|---|
| Announced | September 10, 2026 (API news · deepseek.com news) |
| API model id | deepseek-flash |
| Architecture | 552B MoE; Causal Encoder–Decoder (~8B active input / 16B active output) |
| Context / max output | 1M / 384K |
| Vision / multimodal | Yes (native) |
| Thinking mode | Non-thinking and thinking (default) — see DeepSeek docs |
| Weights | huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash |
API model id and retired names
Set the model to deepseek-flash.
Per DeepSeek’s Sept 10, 2026 announcement:
- V4-Flash and V4-Flash-Vision-Exp are retired. Legacy names
deepseek-v4-flashanddeepseek-v4-flash-vision-exptemporarily route to V4.1-Flash. - DeepSeek is phasing out V4-Pro. Starting 04:00 UTC on September 14, 2026,
deepseek-v4-prorequests route to V4.1-Flash at V4.1-Flash rates until V4.1-Pro launches.
Base URLs (from the pricing page): OpenAI-format api.deepseek.com; Anthropic-format api.deepseek.com/anthropic.
Official API pricing (per 1M tokens)
Prices below are from DeepSeek’s Models & Pricing page for deepseek-flash (DeepSeek-V4.1-Flash). Off-peak rates are 50% of peak.
| Off-peak | Peak | |
|---|---|---|
| Input — cache hit | $0.003 | $0.006 |
| Input — cache miss | $0.15 | $0.3 |
| Output | $0.6 | $1.2 |
Peak hours: 01:00–04:00 and 06:00–10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other hours (including weekends and Chinese public holidays in full) are off-peak.
Other documented features for deepseek-flash: JSON output, tool calls, Responses API, Anthropic API, chat prefix completion (beta), FIM completion (beta, non-thinking only), vision. Concurrency limit listed on the same page: 2500.
New V4.1-Flash pricing took effect at 04:00 UTC on September 10, 2026.
Why DeepSeek positions V4.1-Flash as cheaper to run
From the official announcement (vendor claims, not independently measured here):
- Smaller KV cache vs the previous generation: about 1/4 the HBM and 1/8 the SSD storage — aimed at cutting cache-hit cost for long agent runs.
- Benchmark results described as ahead of flagship models including DeepSeek-V4-Pro (DeepSeek’s own comparison).
Open weights
DeepSeek published the model card and technical report on Hugging Face: deepseek-ai/DeepSeek-V4.1-Flash. The announcement also notes ongoing open-source inference support and large-scale deployment options.
How this page stays accurate
We only cite DeepSeek’s API docs, deepseek.com news, and the official Hugging Face org. Product prices and routing rules can change — re-check pricing and the V4.1-Flash news post before budgeting.
Frequently asked questions
What is DeepSeek 4.1 / DeepSeek-V4.1-Flash?
DeepSeek-V4.1-Flash is DeepSeek’s September 10, 2026 API model: a 552B-parameter MoE with a Causal Encoder–Decoder (about 8B active on input, 16B on output), native vision, 1M context and up to 384K max output. The API model id is deepseek-flash.
What is the DeepSeek-V4.1-Flash API model id?
Use deepseek-flash. Legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp still work but serve V4.1-Flash. From September 14, 2026 04:00 UTC, deepseek-v4-pro also routes to Flash at Flash rates until V4.1-Pro.
How much does DeepSeek-V4.1-Flash / deepseek-flash cost?
Official USD per 1M tokens (off-peak / peak): cache hit $0.003 / $0.006; cache miss $0.15 / $0.3; output $0.6 / $1.2. Peak hours are 01:00–04:00 and 06:00–10:00 UTC Monday–Friday excluding Chinese public holidays.
Does DeepSeek-V4.1-Flash support vision?
Yes. DeepSeek lists native multimodal / vision support for deepseek-flash on the pricing and news pages.
What happened to DeepSeek V4-Flash and V4-Pro?
V4-Flash and V4-Flash-Vision-Exp are retired (legacy ids route to V4.1-Flash). V4-Pro is being phased out; from Sept 14, 2026 04:00 UTC, deepseek-v4-pro requests route to V4.1-Flash at Flash rates until V4.1-Pro launches.
Where can I download DeepSeek-V4.1-Flash weights?
From the official Hugging Face org: huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash (linked from DeepSeek’s announcement).
What is the context window of deepseek-flash?
DeepSeek’s pricing page lists 1M context length and a maximum output of 384K tokens for deepseek-flash (DeepSeek-V4.1-Flash).