deepseek
DeepSeek: DeepSeek V3.1
deepseek/deepseek-chat-v3.1
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
DeepSeek V3.1 (deepseek/deepseek-chat-v3.1) is a large hybrid text-only model from DeepSeek, described by its model card as having 671B total and 37B activated parameters and supporting thinking and non-thinking modes through chat templates. The card states MIT-licensed weights and post-training aimed at better tool use and agent tasks; the OpenRouter catalog lists a long context window and exposes reasoning, tool, and structured-output parameters. Reported benchmark gains are vendor claims; independent evaluation was not available in this source pack.[1][2]
Strengths
- Hybrid operation: one checkpoint offers a thinking mode for harder reasoning and a faster non-thinking mode, switched via the chat template, according to the model card.[1][2]
- Long advertised context: the model card states 128K tokens and the OpenRouter catalog lists 163,840 tokens.[1][2]
- Agent-oriented features: the card documents tool-call, code-agent, and search-agent templates, and the catalog exposes tools, tool_choice, structured_outputs, and reasoning parameters.[1][2]
Limitations
- Only text input and output are listed; neither source documents image, audio, or other modalities.[1][2]
- Self-reported evidence: all benchmark figures come from the vendor's card and the only third-party signal is OpenRouter's Design Arena listing; validate results on your own workloads.[1][2]
- Context figures differ between sources (128K on the card versus 163,840 on OpenRouter), and the card documents tool calling only for non-thinking mode; confirm both with the provider.[1][2]
Where to use it
- Candidate for agentic coding or tool-use pipelines that can exercise the documented tool-call and reasoning parameters; evaluate on your own stack first.[1][2]
- Reasonable candidate for long-document summarization or question answering, given the large advertised context window.[1][2]
- Could suit applications that want optional deeper reasoning on demand while serving quicker non-thinking replies from the same model.[1][2]
Sources
- [1] OpenRouter catalog · DeepSeek: DeepSeek V3.1Sources checked: Nov 13, 2026
- [2] Model documentation · DeepSeek: DeepSeek V3.1Sources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates. It extends the DeepSeek-V3 base with a two-phase long-context...
Specifications
- Model ID
- deepseek/deepseek-chat-v3.1
- Creator
- deepseek
- Added to OpenRouter
- 21.08.2025
- Context length
- 163,840
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.250 / out $0.950
- Knowledge cutoff
- 31.03.2025
- 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 · parameters: include_reasoning, reasoning |
| Coding | No measurement | — |
| Agents | No measurement | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · parameters: include_reasoning, reasoning |
| 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 |
|---|---|---|---|---|
| Design Arena | models · 3dwin rate 47.9% · rank 108 | 1,060Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · codecategorieswin rate 47.8% · rank 107 | 1,102Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · datavizwin rate 46.2% · rank 104 | 1,094Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · gamedevwin rate 47.1% · rank 107 | 1,074Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · svgwin rate 38.2% · rank 97 | 957Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · uicomponentwin rate 47.1% · rank 108 | 1,076Elo | CC — below the B band | 13.11.2026 |
| Design Arena | models · websitewin rate 48% · rank 107 | 1,115Elo | CC — below the B band | 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) |
|---|---|---|---|---|---|---|
| DeepInfrafp4 | 163,840 | $0.250 | $0.950 | — | — | 100% |
| SiliconFlowfp8 | 163,840 | $0.270 | $1 | — | — | 100% |
| CoreWeavefp8 | 161,000 | $0.500 | $1.5 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.