deepseek
DeepSeek: DeepSeek V4 Flash 0423
deepseek/deepseek-v4-flash
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
DeepSeek V4 Flash (listed as deepseek/deepseek-v4-flash) is an efficiency-oriented Mixture-of-Experts text model from DeepSeek with 284B total parameters, about 13B activated per token, and a context window of roughly one million tokens. The provider describes a hybrid attention design intended to cut long-context inference cost, plus selectable reasoning effort modes. It handles text input and output only, and the provider releases the weights under the MIT license. Vendor evaluations look competitive for the size class, but evidence here is limited to the provider card and one aggregator catalog, so independent validation on your own workload is advisable.[1][2]
Strengths
- Context window of about one million tokens, which is unusual for an efficiency-focused model, according to the provider card and the catalog listing.[1][2]
- Sparse MoE design activates only a small share of parameters per token, which the provider positions for fast, low-cost inference.[1][2]
- Selectable reasoning effort modes let callers trade speed for deeper analysis; the vendor reports the highest-effort mode approaches the larger Pro variant on reasoning tasks.[2]
Limitations
- Text input and output only; no other modalities are listed in the catalog.[1]
- The vendor itself states this smaller variant trails the Pro version on pure knowledge tasks and the most complex agentic workflows.[2]
- Evidence is limited to the provider's own card and one aggregator catalog; reported benchmarks and arena results are source-reported or third-party measurements that should be validated on your workload before adoption.[1][2]
Where to use it
- Long-document summarization, search, and analysis that can take advantage of the roughly one-million-token context window.[1][2]
- Coding and reasoning assistants where adjustable thinking effort could balance response speed against accuracy.[1][2]
- High-throughput text pipelines where MIT-licensed weights and a small activated-parameter footprint may favor efficient or self-managed deployment.[2]
Sources
- [1] OpenRouter catalog · DeepSeek: DeepSeek V4 Flash 0423Sources checked: Nov 13, 2026
- [2] Model documentation · DeepSeek: DeepSeek V4 Flash 0423Sources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts model from DeepSeek with 284B total parameters and 13B activated parameters, supporting a 1M-token context window. It is designed for fast inference and...
Specifications
- Model ID
- deepseek/deepseek-v4-flash
- Canonical ID
- deepseek/deepseek-v4-flash-20260423
- Creator
- deepseek
- Added to OpenRouter
- 24.04.2026
- Context length
- 1,048,576
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.090 / out $0.180
- 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: xhigh, high, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Coding | BB — next 30% of the cohort | — |
| Agents | BB — next 30% of the cohort | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · default high · efforts: xhigh, high, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Coding index | 56.2 | BB — next 30% of the cohort | 13.11.2026 |
| Artificial Analysis | Agentic index | 22.2 | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · 3dwin rate 49.3% · rank 62 | 1,171Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · asciiartwin rate 42.7% · rank 61 | 1,105Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · codecategorieswin rate 48.9% · rank 65 | 1,196Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · datavizwin rate 40.7% · rank 94 | 1,125Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · gamedevwin rate 50.2% · rank 62 | 1,187Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · svgwin rate 48.4% · rank 49 | 1,147Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · uicomponentwin rate 44.7% · rank 80 | 1,155Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · websitewin rate 49.1% · rank 66 | 1,200Elo | BB — next 30% of the cohort | 13.11.2026 |
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) |
|---|---|---|---|---|---|---|
| Relacefp4 | 1,048,576 | $0.050 | $0.450 | — | — | 100% |
| StreamLakefp8 | 1,024,000 | $0.084 | $0.168 | — | — | 100% |
| DeepInfrafp8 | 1,048,576 | $0.090 | $0.180 | — | — | 100% |
| GMICloudfp8 | 1,048,575 | $0.091 | $0.182 | — | — | 100% |
| Venice | 1,000,000 | $0.097 | $0.193 | — | — | 100% |
| Fireworks | 1,048,576 | $0.105 | $0.210 | — | — | 80% |
| Wafermxfp4 | 1,048,576 | $0.120 | $0.170 | — | — | 100% |
| SiliconFlowfp8 | 1,048,576 | $0.130 | $0.280 | — | — | 100% |
| Baidufp8 | 1,048,576 | $0.140 | $0.280 | — | — | 0% |
| AtlasCloudfp4 | 1,048,576 | $0.140 | $0.280 | — | — | 100% |
| Parasailfp8 | 1,048,576 | $0.140 | $0.280 | — | — | 75% |
| Mancer 2fp8 | 1,048,576 | $0.150 | $0.500 | — | — | 100% |
| Azure | 1,048,576 | $0.210 | $0.560 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.