 |
In March, Amazon Net Providers (AWS) turned the first cloud service supplier to ship DeepSeek-R1 in a serverless manner by launching it as a completely managed, usually accessible mannequin in Amazon Bedrock. Since then, clients have used DeepSeek-R1’s capabilities by way of Amazon Bedrock to construct generative AI purposes, benefiting from the Bedrock’s strong guardrails and complete tooling for secure AI deployment.
At the moment, I’m excited to announce DeepSeek-V3.1 is now accessible as a completely managed basis mannequin in Amazon Bedrock. DeepSeek-V3.1 is a hybrid open weight mannequin that switches between pondering mode (chain-of-thought reasoning) for detailed step-by-step evaluation and non-thinking mode (direct solutions) for quicker responses.
In keeping with DeepSeek, the pondering mode of DeepSeek-V3.1 achieves comparable reply high quality with higher outcomes, stronger multi-step reasoning for advanced search duties, and large beneficial properties in pondering effectivity in contrast with DeepSeek-R1-0528.
| Benchmarks | DeepSeek-V3.1 | DeepSeek-R1-0528 |
|---|---|---|
| Browsecomp | 30.0 | 8.9 |
| Browsecomp_zh | 49.2 | 35.7 |
| HLE | 29.8 | 24.8 |
| xbench-DeepSearch | 71.2 | 55.0 |
| Frames | 83.7 | 82.0 |
| SimpleQA | 93.4 | 92.3 |
| Seal0 | 42.6 | 29.7 |
| SWE-bench Verified | 66.0 | 44.6 |
| SWE-bench Multilingual | 54.5 | 30.5 |
| Terminal-Bench | 31.3 | 5.7 |
DeepSeek-V3.1 mannequin efficiency in instrument utilization and agent duties has considerably improved by way of post-training optimization in comparison with earlier DeepSeek fashions. DeepSeek-V3.1 additionally helps over 100 languages with near-native proficiency, together with considerably improved functionality in low-resource languages missing massive monolingual or parallel corpora. You’ll be able to construct world purposes to ship enhanced accuracy and lowered hallucinations in comparison with earlier DeepSeek fashions, whereas sustaining visibility into its decision-making course of.
Listed here are your key use circumstances utilizing this mannequin:
- Code era – DeepSeek-V3.1 excels in coding duties with enhancements in software program engineering benchmarks and code agent capabilities, making it excellent for automated code era, debugging, and software program engineering workflows. It performs nicely on coding benchmarks whereas delivering high-quality outcomes effectively.
- Agentic AI instruments – The mannequin options enhanced instrument calling by way of post-training optimization, making it sturdy in instrument utilization and agentic workflows. It helps structured instrument calling, code brokers, and search brokers, positioning it as a strong alternative for constructing autonomous AI programs.
- Enterprise purposes – DeepSeek fashions are built-in into numerous chat platforms and productiveness instruments, enhancing consumer interactions and supporting customer support workflows. The mannequin’s multilingual capabilities and cultural sensitivity make it appropriate for world enterprise purposes.
As I discussed in my earlier put up, when implementing publicly accessible fashions, give cautious consideration to knowledge privateness necessities when implementing in your manufacturing environments, verify for bias in output, and monitor your outcomes by way of knowledge safety, accountable AI, and mannequin analysis.
You’ll be able to entry the enterprise-grade security measures of Amazon Bedrock and implement safeguards personalized to your software necessities and accountable AI insurance policies with Amazon Bedrock Guardrails. You may also consider and evaluate fashions to determine the optimum mannequin on your use circumstances by utilizing Amazon Bedrock mannequin analysis instruments.
Get began with the DeepSeek-V3.1 mannequin in Amazon Bedrock
To check the DeepSeek-V3.1 mannequin in Amazon Bedrock console, select Chat/Textual content below Playgrounds within the left menu pane. Then select Choose mannequin within the higher left, and choose DeepSeek because the class and DeepSeek-V3.1 because the mannequin. Then select Apply.
Utilizing the chosen DeepSeek-V3.1 mannequin, I run the next immediate instance about technical structure resolution.
Define the high-level structure for a scalable URL shortener service like bit.ly. Talk about key parts like API design, database alternative (SQL vs. NoSQL), how the redirect mechanism works, and the way you'll generate distinctive brief codes.
You’ll be able to flip the pondering on and off by toggling Mannequin reasoning mode to generate a response’s chain of thought previous to the ultimate conclusion.
You may also entry the mannequin utilizing the AWS Command Line Interface (AWS CLI) and AWS SDK. This mannequin helps each the InvokeModel and Converse API. You’ll be able to take a look at a broad vary of code examples for a number of use circumstances and a wide range of programming languages.
To be taught extra, go to DeepSeek mannequin inference parameters and responses within the AWS documentation.
Now accessible
DeepSeek-V3.1 is now accessible within the US West (Oregon), Asia Pacific (Tokyo), Asia Pacific (Mumbai), Europe (London), and Europe (Stockholm) AWS Areas. Verify the full Area checklist for future updates. To be taught extra, take a look at the DeepSeek in Amazon Bedrock product web page and the Amazon Bedrock pricing web page.
Give the DeepSeek-V3.1 mannequin a strive within the Amazon Bedrock console in the present day and ship suggestions to AWS re:Submit for Amazon Bedrock or by way of your typical AWS Help contacts.
— Channy
Up to date on September 19, 2025 — Eliminated the mannequin entry part. Amazon Bedrock will simplify entry to all serverless basis fashions, and any new fashions, by robotically enabling them for each AWS account, eliminating the necessity to manually activate entry by way of the Bedrock console. The mannequin entry web page will probably be retired in October 8, 2025 Account directors retain full management over mannequin entry by way of AWS IAM insurance policies and Service Management Insurance policies (SCPs) to limit mannequin entry as wanted.