DeepSeek R1 0528

DeepSeek R1 0528 is a May 2025 refresh of the original DeepSeek R1 reasoning model, positioned as an open-weights alternative to leading closed systems. The release notes highlight improved benchmark performance, stronger front-end generation, fewer hallucinations, and newly added JSON output and function calling, while keeping the same API contract. The model is delivered with openly published weights and is widely routed through third-party providers, making it attractive for teams that want frontier-style reasoning without lock-in to a single vendor.

Under the hood, R1 0528 uses a Mixture-of-Experts design with 671B total parameters and 37B active per inference pass, which is why it can be efficient at serving time despite its scale. It supports a long context window of roughly 164K tokens, handles text input and output, and exposes thinking tokens as part of its open reasoning surface. Together, those traits make it a practical fit for complex analytical tasks, agentic pipelines that need structured outputs or tool use, and research or product teams that want to self-host a strong reasoning model.

Meganovadeepseek-ai/DeepSeek-R1-0528deepseek-thinking

Open provider page Provider docs

Quick Info

Provider: Meganova
Model key: deepseek-ai/DeepSeek-R1-0528
Release date: May 28, 2025
Last updated: May 28, 2025
Knowledge cutoff: 2024-07
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.50
Output token cost: $2.15

Limits

Output tokens: 64,000 tokens
Context window: 163,840 tokens

Transparent token rates

Compare DeepSeek R1 0528 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about DeepSeek R1 0528

Meganova

CoverageAnalysis

Runpod's deep-dive analysis describes DeepSeek-R1-0528 as an update to the DeepSeek R1 reasoning model that retains the Mixture-of-Experts architecture and 128k context length while substantially scaling capabilities. The post reports that AIME 2025 accuracy rose from 70% in the prior version to 87.5% in 0528, with ave The article frames the release as a "minor upgrade" in versioning but a meaningful leap in mathematical and complex reasoning performance attributable to enhanced thinking depth. It characterizes the upgrade as stealthy and low-fanfare, consistent with an incremental open-source release rather than a major architectura

Meganova

CoverageRelease Notes

Opper AI's DeepSeek release tracker catalogues dated model releases and lists DeepSeek R1 (0528) as released on 28 May 2025 with a 164K context window, alongside the smaller R1 0528 Qwen3 8B distilled variant. Pricing entries of $0.50/M input tokens and $2.15/M output tokens appear alongside the R1 (0528) listing, refl As a dated catalog entry, the tracker confirms the 28 May 2025 release date of R1 0528 and its 164K context window, but contains no architecture, benchmark, or capability detail specific to the model. It is useful only for release dating and basic specification context; the prices shown are platform-specific to Opper a

Meganova

Coverage

The official DeepSeek API changelog confirms that DeepSeek-V4-Pro and V4-Flash shipped on 2026-04-24 via both the OpenAI ChatCompletions and Anthropic-compatible interfaces, and that the legacy model aliases deepseek-chat and deepseek-reasoner — which historically mapped to R1-family thinking and non-thinking modes inc Earlier entries in the same changelog show the upstream reasoning lineage that R1-0528 belongs to: V3.1 (2025-08-21) introduced a hybrid reasoning architecture, V3.1-Terminus (2025-09-22) addressed language-consistency and Code/Search Agent issues, V3.2-Exp arrived 2025-09-29, and V3.2 became the default on 2025-12-01.