DeepSeek Releases V4 Pro 0813 With 1.05M Token Context Window, Priced at $0.43/M Input
DeepSeek has shipped the general availability release of DeepSeek V4 Pro, codenamed 0813, featuring a 1,049,000-token context window. The mixture-of-experts model is priced at $0.43 per million input tokens and $0.87 per million output tokens, and is live now on OpenRouter.
DeepSeek V4 Pro 0813 at a glance
DeepSeek Ships GA Release of V4 Pro
DeepSeek has released DeepSeek V4 Pro 0813, the general availability version of its V4 Pro model line. The headline feature is a 1,049,000-token context window, one of the largest context lengths available in a commercially accessible model to date.
The model is a large-scale mixture-of-experts (MoE) architecture, according to DeepSeek, though the company has not disclosed total or active parameter counts. Modality is text-to-text only — no image, audio, or video input/output support has been announced for this release.
Pricing
DeepSeek V4 Pro 0813 is priced at:
- $0.43 per million input tokens
- $0.87 per million output tokens
This pricing positions the model well below frontier-tier offerings from OpenAI and Anthropic on a per-token basis, though DeepSeek has not published benchmark comparisons to substantiate performance parity with those models. No independent benchmark scores (MMLU, HumanEval, or otherwise) were included in this release, and none are yet available from third-party evaluators.
Availability
The model is live now via OpenRouter's API under the identifier deepseek/deepseek-v4-pro-0813. This marks the GA (general availability) status of the V4 Pro line, implying an earlier preview or beta phase preceded this release, though DeepSeek has not detailed what changed between preview and GA builds.
DeepSeek has not disclosed a training data cutoff date, safety evaluation details, or architectural specifics such as the number of expert layers or routing mechanism used in the MoE design. These remain unconfirmed pending further documentation from the company.
What this means
A 1,049,000-token context window is large enough to process entire codebases, lengthy legal documents, or multi-hundred-page technical manuals in a single prompt without chunking. Combined with sub-$1/M-token pricing on both input and output, DeepSeek V4 Pro 0813 targets high-volume, long-context use cases — document analysis, codebase-wide refactoring, and retrieval-heavy agent workflows — where cost per token compounds quickly at scale.
The absence of published benchmark scores is notable. Buyers evaluating this model against GPT-5-class or Claude-class alternatives will need to run their own evals before committing production traffic, since DeepSeek has not provided independent verification of quality at this context length. Long-context models frequently show degraded recall and reasoning quality as prompt length approaches the stated maximum, a pattern seen across the industry regardless of vendor claims.
For teams already running DeepSeek models in production, the 0813 GA tag suggests this is now the stable, supported version of V4 Pro — a signal that earlier preview builds should be migrated off in favor of this release. Given the pricing and context window, this model is likely to see rapid adoption in cost-sensitive, long-document pipelines, provided real-world quality testing confirms DeepSeek's implicit performance claims.
Related Articles
Google releases Gemini 4 Neon at $3/$15 per 1M tokens with a 1M-token context window
Google released Gemini 4 Neon, available today through the API at $3 per million input tokens and $15 per million output tokens. It is the fourth new Gemini 4 model in four weeks, with a context window of about one million tokens.
Thinking Machines Lab releases Inkling 2 Mini: 26B-parameter looped MoE with 3B active and 1M-token context
Thinking Machines Lab has released Inkling 2 Mini, a multimodal Mixture-of-Experts model with about 26B physical total parameters and 3B active per token. It repeats its middle layers three times, supports a 1M-token context, and ships with a DSpark drafter. Hugging Face reports it leads 32 of 48 benchmarks against four comparable models.
Tencent releases Hy4 flagship weights: 770B-parameter MoE, 1,048,576-token context, conditional license
Tencent released downloadable weights for its flagship Hy4 on November 20, with a System Card. The multimodal MoE has a 770-billion-parameter language backbone, 49 billion active parameters per text token and a 1,048,576-token context window, under a license with commercial conditions for model-as-a-service operators.
DeepSeek removes separate fees for model-triggered search and web fetch from Nov 18
DeepSeek removed separate charges for server-side search and web fetch triggered during model calls, effective at noon on November 18, 2026, Beijing time. Token usage is still billed, and the standalone Search and Fetch APIs remain at 5 yuan per 1,000 successful calls.
Comments
Loading...