● Covered by 3 sources · 6 reportsMedium impact6 neutral
DeepSeek Updates V4-Flash Model, Raises API Prices, and Launches V4-Pro
Key points
- DeepSeek-V4-Flash-0731 released in public beta with post-training improvements.
- V4-Flash-0731 uses the same architecture and size as its preview.
- DeepSeek-V4-Pro model launched with agent upgrades and flexible reasoning.
- New API pricing with peak/off-peak rates takes effect August 16, 2026.
- API prices for V4-Pro and V4-Flash will increase significantly.
DeepSeek-V4-Flash-0731 Enters Public Beta
DeepSeek has released DeepSeek-V4-Flash-0731 into public beta, making it available via API. This updated model maintains the same architecture and size as its preview version, with 284 billion total parameters and 13 billion activated parameters per token. The performance gains are attributed to additional post-training.
The DeepSeek-V4-Flash-0731 model has shown improved scores on various benchmarks, including Terminal Bench 2.1 (82.7), NL2Repo (54.2), Cybergym (76.7), DeepSWE (54.4), and Toolathlon verified (70.3). It also scored 89.0% on ARC-AGI-1 Semi-Private and 61.4% on ARC-AGI-2 Semi-Private.
DeepSeek-V4-Pro Launch and Agent Upgrades
DeepSeek has also launched its DeepSeek-V4-Pro model, which includes major agent upgrades and flexible reasoning effort. The V4-Pro model is available on the DeepSeek app/web via "Expert Mode" and through its API. The official release of DeepSeek-V4-Pro was announced alongside the V4-Flash update.
API Pricing Structure Changes
DeepSeek is updating its API pricing with the release of the V4 lineup, introducing peak and off-peak rates. These new rates will take effect at 16:00 UTC on August 16, 2026. Off-peak rates will be 50% lower than peak rates.
The new pricing will significantly increase costs for both V4-Pro and V4-Flash models. For V4-Pro, the peak hour cost for 1 million output tokens will be $3.96, up from $0.87. For V4-Flash, the peak hour cost for 1 million output tokens will be $1.32, up from $0.28. DeepSeek stated that the previous discounted prices were promotional and were initially planned to end on May 31, though they were later made permanent before this new price hike.
Performance and Cost Implications
The DeepSeek-V4-Flash-0731 model is documented to outperform the V4-Pro preview on coding and agentic benchmarks, despite its smaller size and lower price point. The V4-Flash model is priced at $0.14 per million input tokens and $0.28 per million output tokens, while V4-Pro is priced at $0.435 and $0.87, respectively, before the new pricing takes effect.
The company's decision to release the production-ready weights for V4-Flash-0731 under an MIT license on Hugging Face provides organizations with control over deployment and customization.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
Primary sources
GitHub deepseek-ai/awesome-deepseek-integration arXiv 2606.19348 GitHub anomalyco/opencodeHow outlets covered it
DeepSeek is raising the API pricing for its V4 Pro and V4 Flash AI models by four times, effective August 16, 2024. This change introduces peak and off-peak pricing tiers, with the company stating the move is to allocate resources more reasonably, impacting developers and businesses using DeepSeek's AI services.
DeepSeek has released its DeepSeek-V4-Pro model, featuring agent upgrades and flexible reasoning effort. Alongside the release, the company introduced new API pricing with distinct peak and off-peak rates, where off-peak rates are 50% lower.
DeepSeek's refreshed V4-Flash model, priced significantly lower than V4-Pro, is documented to outperform V4-Pro in coding and agentic benchmarks. An analysis was conducted to compare the actual performance and cost of both models on professional coding tasks. The findings aim to clarify why both models exist given the stated performance and price differences.
DeepSeek released its V4 Flash 0731 model, which scored 89.0% on the ARC-AGI-1 Semi-Private benchmark at a cost of $0.02 per task. The model also achieved 61.4% on ARC-AGI-2 Semi-Private at $0.04 per task, indicating its performance on specific AI reasoning tasks.
DeepSeek released DeepSeek-V4-Flash-0731, an updated version of its smaller model, which now surpasses the performance of its larger V4-Pro model on several agent-focused benchmarks through post-training. This development indicates that significant performance gains in AI models can be achieved without increasing model size, offering cost benefits for companies running agents at scale.
DeepSeek has released the DeepSeek-V4-Flash API into public beta, featuring a re-post-trained model (DeepSeek-V4-Flash-0731) with the same architecture and size as its preview. This update primarily affects the DeepSeek-V4-Flash API, with the DeepSeek-V4-Pro API and other models remaining unchanged.