# DeepSeek V4 Flash 0731: Self-Hosting Hardware, Scenarios and Commercial License

> The superseding DeepSeek V4 Flash checkpoint for high-frequency coding, reasoning and agent traffic, with an attached DSpark speculative draft module.

- Verified: 2026-09-12
- Released: 2026-07-31
- Canonical: https://chinaapi.ai/open-model-deployment/deepseek-v4-flash-0731/
- Back to directory: https://chinaapi.ai/open-model-deployment/
- Machine-readable dataset: https://chinaapi.ai/data/open-model-deployment.json

## Model facts

- Vendor: DeepSeek
- Parameters: 304B checkpoint metadata (284B target model plus attached DSpark draft module) total; 13B target path; DSpark draft module attached active
- Native precision: MXFP4 routed experts with FP8/BF16 components; attached DSpark draft weights
- Context: 1M tokens
- Category: General and agentic foundation models
- Modalities: text → text
- Serving paths: Transformers, vLLM, SGLang
- Official weights: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment
- Model documentation: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment

## Deployment evidence

- **Official launch / minimum — evidence B:** The verified vLLM recipe serves the checkpoint on one four-GPU GB300 node at TP4. This is a framework-validated launch shape, not a throughput or concurrency result. Source: https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment
- **200-person team — evidence B:** Start with one measured TP4 GB300 worker and replay the defined 200-seat context, concurrency and reasoning mix; add an independent worker when maintenance or failure recovery is required. Source: https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment
- **Commercial API — evidence B:** vLLM publishes a verified four-GPU GB300 recipe and the vendor documents SGLang TP4. A commercial service still needs independent replicas, admission control, pinned encoding and speculative-decoding validation. Source: https://recipes.vllm.ai/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment

## Scenario evidence

- **High-frequency coding agents — Public evidence:** The 0731 card supersedes the preview checkpoint and reports coding and agent evaluations with the attached DSpark draft module. The scores remain vendor evaluations, not ChinaAPI reproductions. Source: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment
- **Million-token analysis — Vendor-stated:** The target model retains a one-million-token context and a 284B/13B-active MoE shape; the published 304B checkpoint metadata also counts the attached DSpark module. Source: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment
- **Throughput-sensitive private endpoint — ChinaAPI inference:** The official vLLM and SGLang launch paths make the superseding checkpoint a deployment candidate where concurrency matters more than the V4 Pro quality ceiling.

## Commercial-use check

- License: MIT — https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731/blob/main/LICENSE?utm_source=chinaapi&utm_medium=research&utm_campaign=open-model-deployment
- MaaS / hosted service: No model-specific MaaS restriction identified in the MIT license.
- Attribution: Retain the copyright and permission notice.

## Avoid or validate first

- Treating Flash as quality-equivalent to Pro on every task
- Counting 13B active parameters as the checkpoint memory requirement
- Launching a public API without measured parser, speculative-decoding and long-context behavior

## Known limitations and open questions

- No ChinaAPI hardware reproduction or traffic benchmark
- The 304B checkpoint metadata includes the attached draft module and must not be compared directly with the target model's 284B architecture
- The release uses a dedicated encoding implementation rather than a Jinja chat template
- Reasoning effort, parser behavior and DSpark acceptance require workload-specific validation

## Evidence boundary

Loading weights, completing a first forward pass and meeting a production latency/SLA are separate thresholds. The 200-person and commercial tiers require workload-specific measurement. License summaries are product research, not legal advice.

Generated by GENERATED BY scripts/gen_open_model_deployment.py.