Chinese AI lab DeepSeek has released the official DeepSeek-V4-Flash API in public beta. To use the latest version, callers set the model name to deepseek-v4-flash.
The release comes hours after OpenAI launched a price war by discounting its GPT-5.6 Luna. OpenAI cut input token prices from $1 to $0.2 per 1 million, and output tokens from $6 to $1.20 per 1 million, claiming the steep discount was possible after extracting additional architectural efficiencies.
80% discount
OpenAI reduced GPT-5.6 Luna input prices by up to 80 percent.
DeepSeek priced V4 Flash 0731 at $0.14 per 1 million input tokens and $0.28 per 1 million output tokens. The model has 284 billion parameters and offers performance similar to Anthropic's Opus 4.8, which is widely believed to span multi-trillion parameters.
For Code Agent tasks in public benchmarks, DeepSeek tested V4-Flash using the DeepSeek Harness minimal mode with max effort level, topp=0.95, and temperature=1.0. The company also uses internal test sets DSBench-FullStack for full-stack development and DSBench-Hard for hard coding agent problems.
The DeepSeek API now supports V4-Pro and V4-Flash via both the OpenAI ChatCompletions interface and the Anthropic interface. The base_url remains unchanged; callers set the model parameter to deepseek-v4-pro or deepseek-v4-flash.
DeepSeek announced that legacy API model names deepseek-chat and deepseek-reasoner will be discontinued in three months, on 2026-07-24. During the current period, these names point to the non-thinking and thinking modes of deepseek-v4-flash, respectively.
In related moves, DeepSeek upgraded both deepseek-chat and deepseek-reasoner to DeepSeek-V3.2. A temporary endpoint serves DeepSeek-V3.2-Speciale at base_url='https://api.deepseek.com/v3.2_speciale_expires_on_20251215', with the same pricing as V3.2, no tool calls, and availability until Dec 15th, 2025, 15:59 UTC.
Earlier, both models were upgraded to DeepSeek-V3.2-Exp and then to DeepSeek-V3.1-Terminus, which maintained original capabilities while addressing user-reported issues. DeepSeek also upgraded deepseek-chat to DeepSeek-V3-0324 and deepseek-reasoner to DeepSeek-R1-0528, and previously to DeepSeek-V2.5-1210, which optimized file upload and webpage summarization.
The company noted that DeepSeek V2 Chat and DeepSeek Coder V2 were merged into DeepSeek V2.5, accessible via deepseek-coder or deepseek-chat for backward compatibility. V2.5 improved general and code abilities, and the API adopted hard disk caching to reduce prices further. Earlier benchmarks showed DeepSeek-V2-0628's win rate against GPT-4-0314 in Arena-Hard rose from 41.6% to 68.3%, and IFEval accuracy jumped from 63.9% to 77.6% for V2-0517.
Updates
DeepSeek's V4-Flash has demonstrated performance levels comparable to Anthropic's Claude Opus 4.8 in complex coding and autonomous tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. The model offers a massive price advantage, charging approximately $0.28 for the same output that costs $25 on Claude Opus 4.8. Meanwhile, the competitive landscape has expanded with the release of Google's new Gemini 'flash' models, Meta's Muse Spark 1.1, and SpaceXAI's Grok 4.5.
DeepSeek's V4 Flash has debuted ahead of Anthropic's Claude Opus 4.8 on Arena.ai's front-end coding leaderboard, showing performance levels close to the rival model in complex coding and autonomous software tasks. The model offers a significant cost advantage, charging approximately $0.28 for output that costs $25 on Claude Opus 4.8. Additionally, the market has seen new competitive releases, including Google's three efficiency-focused Gemini 'flash' models, Meta's Muse Spark 1.1, and SpaceXAI's Grok 4.5.
DeepSeek's V4-Flash has demonstrated performance levels close to Anthropic's Claude Opus 4.8 in complex coding tests and even debuted ahead of it on Arena.ai's front-end coding leaderboard. The model offers a significant cost advantage, charging approximately $0.28 for the same output that costs $25 on Claude Opus 4.8. Additionally, unverified reports suggest DeepSeek may be preparing for a potential initial public offering (IPO).
DeepSeek's V4 Flash model reportedly performs close to Anthropic's Claude Opus 4.8 in complex coding tasks and has debuted ahead of it on Arena.ai's front-end coding leaderboard. The model offers significant cost advantages, with DeepSeek charging approximately $0.28 for the same output that costs $25 on Claude Opus 4.8. Additionally, sources suggest the startup is preparing for a potential initial public offering (IPO), following the market impact caused by its R1 model in early 2025.
DeepSeek's V4-Flash model has debuted on Arena.ai's front-end coding leaderboard ahead of Anthropic's Claude Opus 4.8, performing close to the latter's level in complex coding and autonomous software tasks. The model reportedly offers a 99% discount compared to Claude Opus 4.8, costing approximately $0.28 for the same output that costs $25. Additionally, unverified reports from Artificial Analysis estimate the average cost of DeepSeek's V4-Flash at 3 cents per test, while sources suggest the company is preparing for a potential IPO.
DeepSeek's V4-Flash has demonstrated performance levels nearing Anthropic's Claude Opus 4.8 in complex coding and autonomous tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. The model offers significant cost advantages, priced at approximately $0.28 for the same output that costs $25 on Claude Opus 4.8. Additionally, the V4-Flash update improves AI agents' capabilities in tool use and multi-step task execution while gaining compatibility with various coding software interfaces.
DeepSeek's V4-Flash model has demonstrated performance levels close to Anthropic's Claude Opus 4.8 in complex coding and autonomous software tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. While Claude Opus 4.8 costs $25 for the same output, V4-Flash charges approximately $0.28, representing a 99% cost advantage. Additionally, research firm Artificial Analysis suggests V4-Flash runs over 60 times cheaper than OpenAI's GPT-5.6 Sol and roughly 29 times cheaper than Moonshot AI's Kimi K3.
DeepSeek's V4-Flash model has demonstrated performance levels close to Anthropic's Claude Opus 4.8 in complex coding and autonomous software tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. The model provides a significant cost advantage, charging approximately $0.28 for the same output that costs $25 on Claude Opus 4.8. Additionally, Artificial Analysis reports that V4-Flash runs at a cost over 60 times lower than OpenAI's GPT-5.6 Sol and approximately 29 times cheaper than Moonshot AI's Kimi K3.
DeepSeek's V4-Flash model has reportedly achieved performance levels close to Anthropic's Claude Opus 4.8 in complex coding and autonomous software tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. The model provides a significant cost advantage, costing approximately $0.28 for the same output that costs $25 on Claude Opus 4.8. Additionally, the V4-Flash update improves AI agent capabilities in tool use and multi-step tasks while gaining compatibility with various software interfaces and coding tools.
DeepSeek's V4-Flash model has demonstrated performance comparable to Anthropic's Claude Opus 4.8 in complex coding and autonomous software tasks, even debuting ahead on Arena.ai's front-end coding leaderboard. The model offers significant cost advantages, charging approximately $0.28 for the same output that costs $25 on Claude Opus 4.8, representing a 99% discount. Additionally, the V4-Flash 0731 model, which features approximately 304 billion parameters, can be run on a single AMD MI300X GPU in production.
DeepSeek's V4-Flash model has demonstrated performance levels near Anthropic's Claude Opus 4.8 in complex coding and autonomous software tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. The model offers a significant cost advantage, charging approximately $0.28 for the same output that costs $25 on Claude Opus 4.8, and is reported by Artificial Analysis to run at a cost over 60 times lower than OpenAI's GPT-5.6 Sol. Additionally, the V4-Flash 0731 model, which features approximately 304 billion parameters, is MIT-licensed and capable of running on a single AMD MI300X GPU in production.
DeepSeek's V4-Flash model has demonstrated performance levels close to Anthropic's Claude Opus 4.8 in complex coding and autonomous software tasks, even debuting ahead of it on Arena.ai's front-end coding leaderboard. The model offers a significant cost advantage, charging approximately $0.28 for output that costs $25 on Claude Opus 4.8, representing a 99% discount. Additionally, the V4-Flash 0731 version can be run on a single AMD MI300X GPU and reached a weekly global usage volume of 7.22 trillion tokens according to OpenRouter data.
DeepSeek's V4-Flash model, now priced at $0.14 per million input tokens and $0.28 per million output tokens, has triggered a surge in usage—processing 8 trillion tokens on August 1 alone—leading the company to warn of imminent price increases due to infrastructure strain, while its performance on the Artificial Analysis Intelligence Index matched Google's Gemini 3.6 Flash at 50/100, one point behind Meta's Muse Spark 1.1 and Z.AI's GLM-5.2.
