deepseek
DeepSeek: DeepSeek V4.1 Flash
deepseek/deepseek-v4.1-flash
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
DeepSeek V4.1 Flash is a multimodal sparse mixture-of-experts model from DeepSeek that accepts text and images and produces text over a context window of up to one million tokens. According to its model card, it is the first model built on the company's Causal Encoder-Decoder architecture, activating 8B parameters during prefill and 16B during decode for cost efficiency on input-heavy workloads. Vendor-reported evaluations emphasize agentic and coding results, and a controllable reasoning effort setting is supported, though independent verification was not available in this source pack.[1][2]
Strengths
- Vendor documentation reports a roughly one-million-token context window combined with aggressive KV cache compression, positioning it for very long inputs at lower serving cost.[1][2]
- The model card reports strong scores on vendor-run agentic and software-engineering benchmarks, including leading results within its own comparison table; these are vendor measurements, not independent tests.[2]
- It accepts image inputs alongside text and, per the OpenRouter listing and model card, supports tool calling, structured outputs, and an adjustable reasoning effort setting.[1][2]
Limitations
- Nearly all performance evidence is vendor-reported and OpenRouter's benchmark fields were empty in this pack, so results should be validated on your own workload before adoption.[1][2]
- In the vendor's own comparison table it trails the best listed frontier model on several harder reasoning and agentic benchmarks, so claimed leadership is not uniform.[2]
- The release ships no standard Jinja chat template and requires a custom encoding toolkit, adding integration work; the knowledge cutoff is not stated in the sources.[1][2]
Where to use it
- Agentic coding and terminal-based software tasks, suggested by the vendor's benchmark emphasis and reported tool-calling support.[1][2]
- Long-document or repository-scale analysis where the reported million-token context and KV cache efficiency may reduce cost.[2]
- Image-plus-text applications such as document understanding, given the reported vision inputs and vendor multimodal evaluations.[1][2]
Sources
- [1] OpenRouter catalog · DeepSeek: DeepSeek V4.1 FlashSources checked: Nov 13, 2026
- [2] Model documentation · DeepSeek: DeepSeek V4.1 FlashSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
DeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on...
Specifications
- Model ID
- deepseek/deepseek-v4.1-flash
- Canonical ID
- deepseek/deepseek-v4.1-flash-20260910
- Creator
- deepseek
- Added to OpenRouter
- 10.09.2026
- Context length
- 1,048,576
- Modality
- text+image->text
- Input modalities
- text, image
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.300 / out $1.2
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | reasoning: optional · default high · efforts: max, high, low, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Coding | No measurement | — |
| Agents | No measurement | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · default high · efforts: max, high, low, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Baidufp8 | 1,048,576 | $0.051 | $0.203 | — | — | 100% |
| Wafermxfp4 | 1,048,576 | $0.051 | $0.567 | — | — | 100% |
| OpenInferencefp4 | 1,048,576 | $0.070 | $0.800 | — | — | 100% |
| InferenceNetfp8 | 1,040,000 | $0.103 | $0.600 | — | — | 100% |
| InferenceNetfp8 | 1,040,000 | $0.103 | $0.600 | — | — | 100% |
| Morphfp8 | 1,048,576 | $0.104 | $0.583 | — | — | 100% |
| Alibaba | 1,000,000 | $0.105 | $0.420 | — | — | 100% |
| DeepInfrafp8 | 1,048,576 | $0.112 | $0.336 | — | — | 100% |
| River | 1,048,576 | $0.120 | $0.480 | — | — | 100% |
| Relacefp4 | 1,048,576 | $0.120 | $0.500 | — | — | 100% |
| Reka | 1,048,576 | $0.120 | $0.640 | — | — | 100% |
| CoreWeavefp8 | 1,048,576 | $0.130 | $0.650 | — | — | 100% |
| AtlasCloudfp8 | 1,048,576 | $0.135 | $0.540 | — | — | 100% |
| DeepInfrafp8 | 1,048,576 | $0.140 | $0.420 | — | — | 100% |
| StreamLakefp8 | 1,024,000 | $0.147 | $0.588 | — | — | 100% |
| GMICloudfp8 | 1,048,575 | $0.180 | $0.720 | — | — | 100% |
| Wafermxfp4 | 1,048,576 | $0.190 | $0.650 | — | — | 100% |
| Coherefp8 | 1,048,576 | $0.195 | $0.660 | — | — | 100% |
| Novitafp8 | 1,048,576 | $0.195 | $0.780 | — | — | 100% |
| DeepInfrafp8 | 1,048,576 | $0.210 | $0.630 | — | — | 100% |
| Fireworks | 1,048,576 | $0.210 | $0.840 | — | — | 100% |
| Mancer 2fp8 | 1,048,576 | $0.212 | $0.750 | — | — | 100% |
| DigitalOcean | 1,048,576 | $0.225 | $0.900 | — | — | 100% |
| Makorafp8 | 1,048,576 | $0.260 | $1.17 | — | — | 100% |
| Crusoefp8 | 1,048,576 | $0.290 | $1.2 | — | — | 100% |
| 1,048,576 | $0.300 | $1.2 | — | — | 100% | |
| BaseTen | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| Together | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| SiliconFlowfp8 | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| Modal | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| BaseTen | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| Parasailfp8 | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| Fireworks | 1,048,576 | $0.300 | $1.2 | — | — | 100% |
| Venicefp8 | 1,000,000 | $0.300 | $1.2 | — | — | 100% |
| Parasailfp8 | 1,048,576 | $0.450 | $1.8 | — | — | 100% |
| Fireworks | 1,048,576 | $0.450 | $1.8 | — | — | 100% |
| Venicefp8 | 1,000,000 | $3 | $12 | — | — | — |
Provider data retrieved: 13.11.2026