DeepSeek model product
DeepSeek-V3.1
1 is a hybrid model supporting both thinking and non-thinking modes through different chat templates.
Updated Aug 12, 2026. Default version: DeepSeek-V3.1
LLMBoard score39.9DeepSeek-V3.1
Coverage80%16 benchmark families
Context window131.1KTokens
Official input priceN/AOfficial price unavailable
DeepSeek-V3.1 Specifications
Technical details for the model's default version.
- Version
- DeepSeek-V3.1
- Released
- Jan 10, 2025
- Knowledge cutoff
- Unknown
- Parameters
- 671B
- Context window
- 131.1K
- Max output
- 8.2K
- Inputs
- text
- Outputs
- text
- Open weights
- Yes
- License
- MIT
DeepSeek-V3.1 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
DeepSeek-V3.1 Benchmark Results
Benchmark scores for DeepSeek-V3.1.
| SimpleQA | 93.4% | 03 | 46 | 95.6% | C | |
| Aider-Polyglot | 68.4% | 08 | 22 | 66.7% | C | |
| BrowseComp-zh | 49.2% | 10 | 13 | 25.0% | C | |
| CodeForces | 69.7% | 12 | 16 | 26.7% | C | |
| Terminal-Bench | 31.3% | 18 | 25 | 29.2% | C | |
| MMLU-Redux | 91.8% | 21 | 48 | 57.5% | C | |
| SWE-bench Multilingual | 54.5% | 29 | 34 | 15.2% | C | |
| MMLU-Pro | 83.7% | 30 | 129 | 77.3% | C | |
| HMMT 2025 | 33.5% | 32 | 33 | 3.1% | C | |
| LiveCodeBench | 56.4% | 36 | 73 | 51.4% | C | |
| AIME 2024 | 66.3% | 43 | 53 | 19.2% | C | |
| BrowseComp | 30.0% | 55 | 58 | 5.3% | C | |
| Humanity's Last Exam | 15.9% | 69 | 93 | 26.1% | C | |
| SWE-Bench Verified | 66.0% | 72 | 105 | 31.7% | C | |
| AIME 2025 | 49.8% | 100 | 114 | 12.4% | C | |
| GPQA | 74.9% | 109 | 234 | 53.6% | C |
DeepSeek-V3.1 Arena Results
Preference and agent-evaluation results for the default version.
| text | overall | 97 | 1419.6 | 3,469 | N/A | |
| text | overall | 100 | 1419.2 | 15,027 | N/A | |
| text | overall | 107 | 1416.5 | 3,702 | N/A | |
| text | overall | 108 | 1416.5 | 11,799 | N/A | |
| text style control | overall | 113 | 1417.2 | 15,027 | N/A | |
| text style control | overall | 114 | 1417.1 | 3,469 | N/A | |
| text style control | overall | 115 | 1416.4 | 11,799 | N/A | |
| text style control | overall | 119 | 1414.5 | 3,702 | N/A |
DeepSeek-V3.1 Runtime Performance
Provider-specific output speed and catalog latency for DeepSeek-V3.1. Runtime does not affect the capability score.
No runtime data
Browse runtime rankingsNo provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
DeepSeek-V3.1 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.20 input, $0.70 output per 1M via NanoGPT
- Tracked offerings
- 19
| NanoGPT | deepseek-ai/DeepSeek-V3.1 | global | $0.20 | $0.70 | 128K | |
| submodel | deepseek-ai/DeepSeek-V3.1 | global | $0.20 | $0.80 | 75K | |
| Vercel AI Gateway | deepseek/deepseek-v3.1 | global | $0.25 | $0.95 | 163.8K | |
| OpenRouter | deepseek/deepseek-chat-v3.1 | global | $0.25 | $0.95 | 163.8K | |
| Deep Infra | deepseek-ai/DeepSeek-V3.1 | global | $0.25 | $0.95 | 163.8K | |
| NovitaAI | deepseek/deepseek-v3.1 | global | $0.27 | $1 | 131.1K | |
| Hugging Face | deepseek-ai/DeepSeek-V3.1 | global | $0.27 | $1 | 131.1K | |
| Jiekou.AI | deepseek/deepseek-v3.1 | global | $0.27 | $1 | 163.8K | |
| Meganova | deepseek-ai/DeepSeek-V3.1 | global | $0.27 | $1 | 164K | |
| Baseten | deepseek-ai/DeepSeek-V3.1 | global | $0.50 | $1.5 | 164K | |
| Merge Gateway | deepseek/deepseek-v3.1 | global | $0.50 | $1.5 | 164K | |
| Weights & Biases | deepseek-ai/DeepSeek-V3.1 | global | $0.55 | $1.65 | 161K | |
| Abacus | deepseek/deepseek-v3.1 | global | $0.55 | $1.66 | 128K | |
| Pioneer | deepseek-ai/DeepSeek-V3.1 | global | $0.56 | $1.68 | 163.8K | |
| Alibaba (China) | deepseek-v3-1 | global | $0.574 | $1.72 | 131.1K | |
| Amazon Bedrock | deepseek.v3-v1:0 | global | $0.58 | $1.68 | 163.8K | |
| Together AI | deepseek-ai/DeepSeek-V3-1 | global | $0.60 | $1.7 | 131.1K | |
| Vertex | deepseek-ai/deepseek-v3.1-maas | global | $0.60 | $1.7 | 163.8K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
DeepSeek-V3.1 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| DeepSeek-V3.1 | 39.9 | 671B | 131.1K | 8.2K | Yes | MIT |
What is DeepSeek-V3.1?
Key information about DeepSeek-V3.1 and its available data.
1 is a hybrid model supporting both thinking and non-thinking modes through different chat templates. 1-Base with a two-phase long context extension (32K phase: 630B tokens, 128K phase: 209B tokens), it features 671B total parameters with 37B activated.
Key improvements include smarter tool calling through post-training optimization, higher thinking efficiency achieving comparable quality to DeepSeek-R1-0528 while responding more quickly, and UE8M0 FP8 scale data format for model weights and activations.
The model excels in both reasoning tasks (thinking mode) and practical applications (non-thinking mode), with particularly strong performance in code agent tasks, math competitions, and search-based problem solving.
Data as of 2026-08-11.
DeepSeek-V3.1 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to DeepSeek-V3.1
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
DE
DeepSeek-R1
DeepSeek
DE
DeepSeek-V3
DeepSeek
DE
DeepSeek-V3.2
DeepSeek
DE
DeepSeek-R1-Distill-Qwen
DeepSeek
DE
DeepSeek-V2.5
DeepSeek
DE
DeepSeek-R1-Distill-Llama
DeepSeek
FAQ
Common questions about DeepSeek-V3.1.
When was DeepSeek-V3.1 released?
DeepSeek-V3.1's default version was released on Jan 10, 2025.
How much does DeepSeek-V3.1 cost?
No official standard PAYG price is currently available for DeepSeek-V3.1. The lowest tracked third-party offer starts at $0.20 input and $0.70 output via NanoGPT.
Who created DeepSeek-V3.1?
DeepSeek-V3.1 was created by DeepSeek.
What is the context window for DeepSeek-V3.1?
The default version has a 131.1K token context window.
Is DeepSeek-V3.1 open weight?
Yes. The default version is marked as open weight under MIT.
How many API providers offer DeepSeek-V3.1?
19 provider offerings are linked to the default version.
What models should I compare DeepSeek-V3.1 with?
Nearby ranked alternatives include Grok 4.3, Qwen3 Max, Qwen3 VL 32B Thinking.