At a Glance
| Vendor | DeepSeek |
| Family | DeepSeek V series |
| Launched | DeepSeek released V4.1-Flash on September 10, 2026, according to its API changelog. It is served on the API as deepseek-flash and replaced V4-Flash. DeepSeek has referred to a V4.1-Pro but has not released one, so V4-Pro, generally available since August 13, 2026, remains its larger model. An earlier version of this page dated V4.1 to June 2026 and listed a V4.1 Pro, which was wrong. |
| Context window | 1,000,000 tokens with a 384,000-token maximum output. The model has 552B parameters, with 8B active on input and 16B on output, and accepts image input. |
| Pricing | Per million tokens, V4.1-Flash costs $0.15 input and $0.60 output off-peak, and $0.30 input and $1.20 output at peak, under the time-of-day billing DeepSeek introduced on August 16, 2026. Cached input costs $0.003 off-peak and $0.006 at peak. Weights are on Hugging Face under the MIT licence. |
| Access channels | DeepSeek API (deepseek-flash), the DeepSeek app and web chat, and open weights on Hugging Face under MIT, which makes it available through third-party hosts and for self-hosting. |
Notable Benchmarks
DeepSeek published a technical report with the release. We have not verified individual benchmark scores for V4.1-Flash against a primary source, so none are quoted here. Trackers cite a Codeforces rating of 3471 from the model card, and we found no published LiveCodeBench score.
Strengths
Very low per-token pricing, a 1M context with a large output limit, image input, and MIT-licensed open weights.
Limitations
There is no V4.1-Pro yet. Peak-hour pricing doubles the rate. Chinese-origin governance concerns persist for some Western buyers in regulated industries.
Brand-Visibility Implications
Because V4.1-Flash is cheap and open-weight, it ends up inside developer tools and self-hosted RAG systems where nobody outside can observe the answers. Brand visibility there depends on what the model learned in training and on documentation that crawlers can fetch: GitHub, technical docs, and well-structured product pages. See DeepSeek visibility, DeepSeek V4 brief, and the DeepSeek model lineage.
How Presenc AI Tracks This Model
Presenc AI monitors brand visibility on DeepSeek's DeepSeek V series as part of continuous multi-platform AI visibility tracking. We sample DeepSeek V4.1 across representative prompt sets daily, compare against competitor performance on the same prompts, and flag material mention-rate changes so brand teams can respond quickly when AI representation shifts.