DeepSeek R1 0528: Self-Hosting & Deployment Guide
May 28th update to the original DeepSeek R1 Performance on par with OpenAI o1, but open-sourced and with fully open reasoning tokens.
What is DeepSeek R1 0528?
May 28th update to the original DeepSeek R1 Performance on par with OpenAI o1, but open-sourced and with fully open reasoning tokens.
Developer: DeepSeek. Released 28 May 2025. Context window 163,840 tokens, up to 32,768 output tokens.
How to run DeepSeek R1 0528 yourself
Open weights (mit), 684.5B parameters. Below: how much memory the weights need at each quantization, which hardware fits them, and the commands to serve it.
DeepSeek R1 0528 VRAM requirements
| Precision / quantization | Weights | Figure | Source |
|---|---|---|---|
| BF16 (full precision) | 1369.1 GB | Estimate | 684.5B params × 2 bytes |
File size = actual download size on Hugging Face. Estimate = parameter count × bytes per weight. Weights only — the KV cache for long contexts needs extra memory; the fit tables below add 15% headroom as a rule of thumb.
GPU requirements: deploy DeepSeek R1 0528 on your hardware
| Hardware | BF16 (full precision) |
|---|---|
| NVIDIA H200141 GB | — |
| NVIDIA H10080 GB | — |
| NVIDIA A100 80GB80 GB | — |
| NVIDIA L40S48 GB | — |
GPUs needed per variant.
| Hardware | BF16 (full precision) |
|---|---|
| AWS p5.48xlarge8× H100 · 640 GB | — |
| AWS p4de.24xlarge8× A100 80GB · 640 GB | — |
| AWS g6e.12xlarge4× L40S · 192 GB | — |
| Google Cloud a3-highgpu-8g8× H100 · 640 GB | — |
| Azure ND H100 v58× H100 · 640 GB | — |
Whole-instance GPU memory; public instance specs.
Serve it yourself
pip install -U vllm
vllm serve deepseek-ai/DeepSeek-R1-0528pip install 'sglang[all]'
python -m sglang.launch_server --model-path deepseek-ai/DeepSeek-R1-0528Commands use the official repositories above; vLLM and SGLang expose an OpenAI-compatible endpoint.
Not on LLM.API yet
We do not route this model through our API at the moment. Browse the models you can call today — most workloads have a close match already live.
Providers
Companies that host this model today, with their public list prices. LLM API does not route this model yet.
List price by provider ($ / 1M tokens)
InputOutputProvider list prices from OpenRouter's public catalogue.
| Provider | Input /M | Output /M | Cache read /M | Context | Precision | Uptime (24h) |
|---|---|---|---|---|---|---|
| DeepInfra | $0.5 | $2.15 | $0.35 | 163,840 | fp4 | 100.0% |
| SiliconFlow | $0.5 | $2.18 | — | 163,840 | fp8 | 99.4% |
| StreamLake | $0.571 | $2.286 | — | 128,000 | — | 99.6% |
| Novita | $0.7 | $2.5 | $0.35 | 163,840 | fp8 | 99.9% |
Source: OpenRouter public catalogue. Last updated 23 Sep 2026.
Not available on LLM API yet
We do not route this model through our API at the moment, so there is no endpoint or code snippet for it yet. Browse the models you can call today — most workloads have a close match already live.
Browse available modelsWhy Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Intelligent Model Routing
Dynamically route each request to the optimal model across providers based on latency, price, and quality—without changing your integration or redeploying code.
One API, many models -
Cost-Aware Orchestration
Automatically balance premium and budget models with per-call controls, caps, and policies so you can ship fast while keeping AI spend predictable and optimized.
Control spend by design -
Resilient Fallback Flows
Define cascading provider and model fallbacks so timeouts, quota limits, or regional outages transparently fail over—maintaining uptime without custom retry logic.
Fail soft, not hard -
End-to-End Observability
Trace every request across providers with logs, metrics, and structured payloads to debug prompts, compare models, and tune performance from a single dashboard.
See every token -
Task-Level Abstractions
Call high-level tasks like chat, generate, extract, or rank instead of vendor-specific APIs, so you can swap models without rewriting business logic.
Code to tasks, not vendors -
High-Throughput Batch Jobs
Run large-scale inference workloads as managed batches with concurrency, retries, and progress tracking built in—perfect for backfills, evaluations, and data processing.
Crush your backlogs
5 Core Capabilities
Deep chain-of-thought reasoning
Post-training added reasoning depth: the model card reports roughly 23K thinking tokens per AIME question, up from about 12K in the previous R1.
Maths and competition problems
AIME 2025 accuracy rose from 70% to 87.5% between the original R1 and this checkpoint.
Lower hallucination rate
The release notes cite a reduced hallucination rate alongside the reasoning gains.
Function calling and JSON output
This checkpoint added JSON output and function-calling support, plus system-prompt support that earlier R1 lacked.
Open weights under MIT
A 671B-parameter mixture-of-experts model with 37B active per token, published openly for self-hosting.
6 Most Valuable Use Cases
- Maths, logic and competition-style problem solving
- Research assistants that must show their working
- Code debugging where a long deliberate trace beats a fast guess
- Structured extraction returning validated JSON
- Tool-calling agents that need explicit multi-step planning
- Self-hosted reasoning workloads under a permissive licence
When to Use — When NOT to Use
Use it if...
- Your task rewards deliberate reasoning over fast replies
- You need strong maths and logic results at open-weight pricing
- You want reasoning traces you can inspect and audit
- You need open weights for self-managed evaluation
Avoid if...
- You are latency-sensitive: longer traces mean more tokens and more waiting
- You are cost-sensitive per answer, since thinking tokens are billed
- You need multimodal input — this model is text only
- You want the current DeepSeek hosted generation, which has moved on from R1
BENCHMARKS
DeepSeek R1 0528 benchmark scores
Intelligence index
Scale: 0-100 index points
Reference price per 1M tokens
Bars compare published reference list prices, not LLM.API pricing.
Scores as published by Artificial Analysis and the official model card (source). Reference prices are provider list prices, not LLM.API pricing. Figures with no published value are omitted.
COMMUNITY
What developers say about DeepSeek R1 0528
Summarised from publicly published developer write-ups and the model's own documentation. Opinions are the sources’, not LLM.API’s.
- Coverage describes R1-0528 as a checkpoint update: same base lineage, a heavier post-training run, and noticeably deeper reasoning.
- The practical changes most often mentioned are system-prompt support, JSON output and function calling, which earlier R1 lacked.
- Reviewers warn that a higher score comes with more generated tokens, so latency and cost per answer rise with it.
- Hosted and open-weight surfaces differ: the hosted release used a 64K context limit while the open checkpoint supports 128K.
SOURCES
Frequently Asked Questions
What is DeepSeek R1 0528?
May 28th update to the original DeepSeek R1 Performance on par with OpenAI o1, but open-sourced and with fully open reasoning tokens.
Who makes DeepSeek R1 0528?
DeepSeek R1 0528 is developed by DeepSeek. It was released on 28 May 2025.
What is the context window of DeepSeek R1 0528?
163,840 tokens, with up to 32,768 output tokens per response.
How much does DeepSeek R1 0528 cost?
The reference list price is $0.5 per 1M input tokens and $2.15 per 1M output tokens. The cheapest host right now is DeepInfra at $0.5 / $2.15 per 1M tokens.
Which providers host DeepSeek R1 0528?
DeepInfra, SiliconFlow, StreamLake, Novita.
What modalities does DeepSeek R1 0528 support?
Text input and text output. It supports tool calling, structured outputs, JSON mode, reasoning.
Can I use DeepSeek R1 0528 through LLM API?
Not yet. LLM API does not route this model at the moment. Browse the models page for close alternatives you can call today with one API key.
COMPARE
Competitive Models
-
DeepSeek V4 Flash
Up to 30%
DeepSeek V4 Flash is an efficiency-optimized Mixture-of-Experts large language model from DeepSeek, featuring a 1M-token context window and fast inference for high-throughput applications.
-
DeepSeek V4 Flash (free)
Up to 30%
DeepSeek V4 Flash (free) is an open-source, efficiency-optimized Mixture-of-Experts language model from DeepSeek, offering a 1M-token context window with only 13B parameters activated per token out of 284B total. It is designed to deliver fast, cost-effective long-context reasoning, coding, and agentic workflows.
-
DeepSeek V3.2
DeepSeek V3.2 is a large open-source Mixture-of-Experts language model from DeepSeek that emphasizes high reasoning performance and efficient long‑context inference. It is notable for its DeepSeek Sparse Attention and multi-latent attention mechanisms, which significantly cut compute and memory costs for long sequences.
Get one key to every model
Swap your API key. Keep your code.