deepseek
DeepSeek: DeepSeek V4 Pro 0423
deepseek/deepseek-v4-pro
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
DeepSeek V4 Pro 0423 is a text-in, text-out Mixture-of-Experts model from DeepSeek listed in the OpenRouter catalog, with 1.6T total and 49B activated parameters and a one-million-token context window. Vendor documentation describes it as a preview release aimed at reasoning, coding, and agentic work, with selectable reasoning effort modes. The vendor's self-reported evaluations suggest strong coding and knowledge results at maximum effort, while supplied arena-style measurements are more mixed. Evidence here is limited to a catalog entry and the vendor's own model card, so independent verification is thin.[1][2]
Strengths
- One-million-token context window; the vendor reports a hybrid attention design that sharply reduces long-context inference compute and KV cache versus its predecessor.[1][2]
- Tunable reasoning: non-think, high, and maximum effort modes, with the OpenRouter catalog also exposing tool calling, tool choice, structured outputs, and reasoning effort parameters.[1][2]
- Vendor-reported benchmark tables place its strongest results in maximum-effort mode on coding and knowledge tasks, though these figures are not independently confirmed.[2]
Limitations
- All headline benchmark numbers are self-reported by DeepSeek; the independent arena measurements supplied are middling rather than leading, so quality on your own workload needs validation.[1][2]
- Evidence is limited to an OpenRouter catalog entry and the vendor's own card: knowledge cutoff is unstated, deployment hardware requirements are not given, and the release is described by the vendor as a preview.[1][2]
- The catalog lists text as the only input and output modality; multimodal capability is unknown rather than confirmed absent.[1]
Where to use it
- Potentially well suited to long-document question answering, large-codebase analysis, or multi-document synthesis that could exploit the million-token window, subject to validation.[1][2]
- A candidate for coding assistants and agentic tool-use pipelines, given catalog support for tools and structured outputs and the vendor's emphasis on coding and agentic evaluations.[1][2]
- May fit workloads that benefit from switching between fast non-reasoning replies and deeper thinking modes within one model.[1][2]
Sources
- [1] OpenRouter catalog · DeepSeek: DeepSeek V4 Pro 0423Sources checked: Nov 13, 2026
- [2] Model documentation · DeepSeek: DeepSeek V4 Pro 0423Sources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding,...
Specifications
- Model ID
- deepseek/deepseek-v4-pro
- Canonical ID
- deepseek/deepseek-v4-pro-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 $0.900 / out $1.5
- 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 | CC — below the B band | reasoning: optional · default high · efforts: xhigh, high, 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: 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 | Intelligence index | 12.6 | CC — below the B band | 13.11.2026 |
| Design Arena | agents · fullstackwin rate 22.1% · rank 45 | 948Elo | CC — below the B band | 13.11.2026 |
| Design Arena | agents · godotgamedevwin rate 34% · rank 32 | 1,059Elo | CC — below the B band | 13.11.2026 |
| Design Arena | agents · webappswin rate 26.4% · rank 46 | 1,000Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · 3dwin rate 56.8% · rank 37 | 1,236Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · asciiartwin rate 46.4% · rank 46 | 1,143Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · codecategorieswin rate 52.2% · rank 50 | 1,232Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · datavizwin rate 48.8% · rank 60 | 1,201Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · gamedevwin rate 53% · rank 46 | 1,222Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · svgwin rate 45.5% · rank 58 | 1,129Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · uicomponentwin rate 50.6% · rank 57 | 1,212Elo | BB — next 30% of the cohort | 13.11.2026 |
| Design Arena | models · websitewin rate 50.9% · rank 57 | 1,228Elo | 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) |
|---|---|---|---|---|---|---|
| Baidufp8 | 1,048,576 | $0.828 | $1.66 | — | — | 100% |
| StreamLakefp8 | 1,024,000 | $0.829 | $1.66 | — | — | 100% |
| Relacefp4 | 1,048,576 | $0.900 | $1.5 | — | — | 100% |
| GMICloudfp8 | 1,048,576 | $0.957 | $1.91 | — | — | 100% |
| SiliconFlowfp8 | 1,048,576 | $1 | $1.5 | — | — | 100% |
| DigitalOcean | 1,048,576 | $1.04 | $2.09 | — | — | 100% |
| DeepInfrafp8 | 1,048,576 | $1.3 | $2.6 | — | — | 100% |
| Alibabafp8 | 1,000,000 | $1.42 | $2.83 | — | — | 100% |
| Alibabafp8 | 1,000,000 | $1.56 | $3.12 | — | — | 100% |
| Novitafp8 | 1,048,576 | $1.6 | $3.2 | — | — | 100% |
| Venice | 1,000,000 | $1.65 | $3.3 | — | — | 100% |
| AtlasCloudfp4 | 1,048,576 | $1.68 | $3.38 | — | — | 100% |
| Azure | 1,048,576 | $1.91 | $3.83 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.